Peter in the Media

Hi everyone, Peter here, CEO and co-founder of Syrvi. Today I am going to talk about Claude's new release, Opus 4.8. It is supposed to be an amazing model and I have been testing it like crazy since last night. As always, my videos are meant for entrepreneurs and founders of small to medium businesses, I would say between 10 and 50 employees.

So why should you be listening to me? As I said, we do outreach, content marketing, lead generation and paid advertising across every platform you can imagine. We have spent and made hundreds of thousands of pounds on paid advertising and outreach for our clients. I think the most important thing for any business is a revenue engine. One, multi-channel outreach: email and LinkedIn. Two, content marketing with paid advertising and brand building, the content engine. Three, nurturing, follow-up and a booking system on autopilot. These are the things that determine how much money you make, so your return on investment is really important.

Why am I doing this video today? Because I and the whole Syrvi team work with multiple AI models. We test across all kinds of LLMs, and we use Claude for most of the stack. So we use a number of LLMs in the back end that power our revenue engine, but it is always Opus, 4.5, 4.7 and now 4.8, which is the most powerful model. Opus is your brain, it can power your revenue engine. I strongly recommend you stick with Claude, it is the most amazing tool out there, especially Opus, and always the newest model.

More importantly, I created something for you: how Syrvi gets the best out of the model for our ecosystem, for our Q2 and Q3 planning, and how we use it for coding too. I might do a separate video on the coding side. You do not need to be a coder and you do not need to be a developer. But I do believe that if you are an owner, founder, C-level, or even a head of marketing or sales, any revenue-generating department, it is really important to be able to solve certain things yourself. Anything that takes no more than a couple of hours, maybe a day or two, is better if you build it yourself: any tool that helps generate revenue, marketing and sales assets, case studies, landing pages, paid-advertising tools, or your own productivity, managing your calendar, a personal assistant, an AI agent. All of this can now be built with Claude Code and Opus, and you do not need to be a developer.

Now, if you have a team and you develop something that you think could be bigger, I always develop my own proof-of-concept products. Whatever I can implement for myself, I do it myself to increase my productivity. And whenever I think this could be a lot bigger, this could actually be a product, this could be part of the revenue engine, but it is way over my technical expertise, I hand it on. The last thing we want, if we are in a certain position with certain experience, is to become developers. Not because there is anything wrong with the job, it just requires a completely separate skill set and mindset, and we are not necessarily going to change our role.

So, long story short, five things that change how we work.

One, effort is in the dial, not the prompt

Opus 4.8 has four effort levels, and the default comes with high effort everywhere. When the reasoning looks shallow on a hard task, raise the effort instead of rewriting the prompt. Planning is still critical, how you plan a prompt and a task, whether simple or multi-step. But rather than going back and forth, you change it to low or medium for low-volume work, high by default, or extra when you do not like the answer. There is even a max setting, which I am testing, it uses more of your limit and more credit, but the output can be significantly better than extra. So rather than playing ping-pong with the agent, you just change the effort.

Two, it is materially more honest

This is very important. I spent a lot of time wrestling with Opus 4.7, and honestly I had some of the biggest arguments with an LLM that anyone has had, because I work with it actively. 4.7 had attitude and sometimes became lazy, and I am doing more complex tasks, developing a whole AI ecosystem, which I will also do a video about. The problem I faced many times was it almost talking back and not doing what it was supposed to do, and worst of all, making the same mistake again. I used to get that from human teams, but I do not expect it from AI. The biggest improvement here is a self-executing ecosystem, which is what I built internally with the team. 4.7 caused us some issues, but 4.8 is a lot better, including at scoring its own work, which 4.7 was really bad at.

Three, dynamic workflows make engine-scale work possible

Claude Code can plan a task, spin up hundreds of parallel sub-agents, verify its own outputs and report back, including across a code base. I will do a separate video for those who want to get more technical, on using Claude Code in the terminal. But you do not have to: you have chat, and you have Cowork, which has better browsing and can access your SharePoint folder and work in it. Chat is just a chat function; Claude Code is where you can code. If you have never coded, I would highly recommend starting in Cowork, connect your SharePoint folder, connect a lot of other things, and start to play around. You can also use the terminal.

Four, fast mode is now cheap

If you switch the model to fast, it is two and a half times the speed at three times cheaper. Use it for high-volume, low-stakes work: content variants, classification, enrichment. Keep standard high effort for strategy and code. Same-string swap: the baseline from Opus 4.7 to 4.8 has no breaking API change, but rebaseline your effort, latency and cost.

Five, it is more honest about its own work

The fact that matters most to us is that it is roughly four times more likely to honestly flag its own uncertainty. In my case, that means less oversight and less babysitting. Honesty is really important: if I have done a job, I expect it to have done its job, rather than telling me it is done when it is not, so that I then have to check, and coming back with four or five decisions that make no sense for me to spend my time on. That is good advice in general: make sure your execution is goal-driven. I do not care what changes it makes, as long as they are reversible, there is a recovery path and history, so it can change things back. If something is irreversible, then of course I care. But otherwise, do not give me four or five options, just decide.

Coding is a different thing, and I will do a coding video in the terminal too, because if I can do it, you can do it. I am genuinely not a coder, with very little technical expertise, but I still need to understand what is possible in order to be a CEO and founder running these businesses and managing teams.

A few more things we discussed: computer use through the browser, agentic enrichment, the effort dial (low, medium, high, and max, which is what we will use most), and dynamic workflows, which are a game changer for coding in the terminal. Claude Opus 4.8 can create hundreds of sub-agents, so at any given time, regardless of how difficult and complex the project is, it can plan the project and assign work to sub-agents. Migrate first, then trust it.

Scheduled tasks

If you have scheduled tasks in Cowork, and you do not need to be a coder for this, you can schedule as many as you want. The good thing is that Cowork does the job on autopilot: LinkedIn engagement, multi-performance reports, a whole ecosystem audit, a continuous second brain, strategic-priority check-ups, in-loop audits, trend signals, content-engine builds, whatever it might be. With most AI you know you have to go there, put in the prompt and trigger the task. With scheduled tasks you plan the prompt once, and it runs every day, multiple times a day, weekly, monthly or quarterly, pulling data from the CRM however you want. It is always there.

CLAUDE.md and goal-driven execution

If you want, I can share a PDF. CLAUDE.md is really important, it is a rule book. If you want to build any AI ecosystem or second brain, like a knowledge base, CLAUDE.md is essentially an SOP for Claude: before you plan or do anything, this is who you are and this is how you do it. This is your working space, this is who I am, this is my company, this is who we are selling to, and this is how you need to behave, plus all the skills and slash commands, which you can change.

This is way more than just using Claude, but I am sharing how we build our ecosystem. For me there are three important goals when using Claude, or honestly any AI tool. This resonates a lot with Andrej Karpathy, who recently joined Anthropic and previously helped at OpenAI, one of the most respected experts in the field. The CLAUDE.md file, which you can download from GitHub, is essentially gold for goal-driven execution by Claude, and that is what I am using too.

So, overall, goal-driven: what is your next quarterly goal or next-year goal? Second, goal-driven task execution: how does any task you are doing contribute to that overall goal? And how does the self-executing, self-learning system help, it needs to learn from its mistakes. It creates daily logs and a drift index, so whenever something fails, or I give feedback, or it comes from analytics, it sees how it can improve, so each task gets better and stays goal-driven. Each task has to contribute to the overall goal.

That is how we do it. Thank you very much for your time. I am going to do more videos around the AI ecosystem, the second brain, the engine, and anything else Claude releases, because we use Claude, and a lot of videos on how to connect with Zoho One, which is our favourite tool. So Claude, Claude Code and Zoho One, a lot of context around that, and more generally about B2B lead generation, especially for UK SMEs, but we also work in the US, Australia, New Zealand and the EU, so I will be tackling a few things over there as well.

What could I do to build something similar, so I can literally execute this, create a plan and execute it within even a day, or two, or three? There are three pillars we are going to build. I am not going to build it live, but I am going to show you the exact steps, and also which tools we would use.

We would use Claude, from the many AI platforms out there. There is OpenAI, there is Google's Gemini models, there is Perplexity, there is so much AI software you can find. We tend to use Claude, and Claude is from Anthropic. You can get a subscription, there is a free tier and there are paid subscriptions, and within Claude we would use the Cowork function. Using Cowork, I will show you exactly how I would build it. From the previous decks, where I showed you the Syrvi revenue engine and how we built it, this is how you would build it yourself if you just got a subscription on Claude, opened Cowork, and went from there.

It is not available from a web browser and is not available on free plans. Cowork is only available on Pro, Teams and Enterprise subscriptions, and you also need the Claude desktop app. Then you can enable Claude Cowork. This is a tool that can help you with strategy, where you can add different plug-ins and connectors, and I will show you exactly why that matters, because these are the steps we will go through: open Cowork from the menu, install the plug-ins, wire up the connectors, and once that is done: ICP creation and research (as I showed you in the plan), competitor research, segmentation, total addressable market, then targeting and channels, and finally data enrichment. The last step is which tools we will use to send the emails and outreach we have developed.

So if you have a Claude Anthropic subscription, you can open Cowork: install the desktop app and switch to Cowork mode. If you use OpenAI or Google, you will most likely have very similar functions. We tend to use Claude because we believe it is the most advanced tool SMEs can benefit from. For different purposes Google Gemini might work better, or for others OpenAI's new ChatGPT models might work better, but generally I believe Claude is perfect for building an AI revenue engine. If you have other models, you will probably find very similar functions.

Skills, connectors and plug-ins

In the chat composer you will see a plus button, where you can select skills, connectors and plug-ins. Skills: you can tell Claude, or any LLM (you can use skills with other models too), a set of proven instructions that teach Claude or any AI model to do a job well. Connectors are a major development in this new world, because before, you needed a technical team and skill set, you had to build automations to connect different tools together, which is not easy or simple. Within Claude you can use connectors without APIs and all the technical work: you just connect with the tools you need to build your AI revenue engine. Plug-ins are nothing else but bundles of skills: you can create multiple skills and bundle them into a plug-in.

Install the marketing and sales plug-ins

The first step: go to the chat menu, the plus button, click on plug-ins, and add the marketing, sales and productivity plug-ins. You can even add design ones, but these three are the most important. Marketing handles everything you need: your ICP, the ideal client profile development, competitor research, segmentation and TAM. Sales handles the outreach itself, call preparation and pipeline. Productivity is the memory layer, so every new Cowork session can build from the previous one; you do not need to build and explain everything from scratch each time. You can use the slash command in the chat and you should see the commands. Once you do this you can see: draft content, competitive brief, campaign plan, draft outreach. You choose whichever function you want once the plug-in is installed.

Wire up your connectors

This is critical. You can use your Claude and Cowork space to connect with third-party tools, which was unimaginable even six months or a year ago: that you can build a system without being a coder or an automation expert. You just connect with these third-party tools through your connectors; all you need is an authentication, an authorised connection and access, and it is available immediately from your Claude Cowork space.

Mapping the engine steps: for ICP creation and research you need a marketing plug-in plus connectors that supercharge it, for example a web search function, Apollo or ZoomInfo. For competitor research, a marketing plug-in and web search, Similarweb or Ahrefs. For segmentation, a marketing plug-in again, Apollo or Clay, and if you have a CRM such as Zoho or HubSpot you can connect that too. For total addressable market estimation, a marketing plug-in, plus connectors like Apollo, ZoomInfo and web search. For targeting angles, a marketing plug-in plus web search, Clay or LinkedIn signals. For channel and sequence orchestration, a sales plug-in, connecting Instantly, HeyReach or Outreach. For data and enrichment, a sales plug-in and a tool connector: Apollo, ZoomInfo, Lusha, Clay. There are so many.

Build the ICP first

Your ICP is the foundation of everything downstream, so the marketing plug-in pulls it together. On your existing customers you can run a web search and synthesise a sharp, evidence-backed ICP. Open a new Cowork chat, drop in five to fifteen of your best customers, each with, for example, their sector and what they bought from you, and then take this prompt and run it in Claude Cowork, or any other model: "You are my GTM strategist. I want a sharp, evidence-backed ICP for XYZ company, which does this and that." Upload a CSV or Excel file with your top ten or fifteen customers, follow the instructions in the prompt, copy and paste it, and it will help develop your ICP, the ideal client profile.

Competitor research is again a marketing plug-in. Use the command, and there is a prompt you can copy and paste. If you have a Similarweb or Ahrefs subscription, connect it with your Cowork or whichever model you use, and you can add SEO signals to the brief, which makes the outreach messaging and angles even more relevant. You can always reuse any Claude template or brief from previous sessions. The prompt is on the right; screenshot it, or request it after the session, happy to share.

Segmentation and TAM

With the ICP developed and the competitive brief in place, ask Cowork to slice the market into three to five actionable segments and size each one. You need to understand your total addressable market and the size by segment. If the connectors are in place, using Apollo or ZoomInfo, and even web search to compare sources, Cowork can estimate the size of each segment, how many companies fall into the client profile you developed. Again you have two prompts: prompt A for segmentation, prompt B for TAM estimation. Screenshot them, or request them after the session.

Targeting angles and channel orchestration

This is where it gets going, because now you have all the intel for your AI agents to write highly resonating, highly targeted copy. Now you know who, and you need how. Two prompts: the first generates targeting angles; the second orchestrates the actual cross-channel sequence, for example cold email with LinkedIn outreach. Prompt A is targeting angles, prompt B is sequence orchestration. Add them to Claude or any other AI model you use, and it will help develop whatever you need.

Data and enrichment

If you have all the connectors, Apollo, ZoomInfo, Lusha, this is the point where you have everything you need to extract the data. To send an email to anyone, or connect with anyone on LinkedIn, you need data, a prospecting list. You will need a subscription with each of these tools separately, but once you have it you can connect it with your Claude or any other AI tool. There is also Clay, which can do the enrichment. Prompt A builds the list: how many companies and prospects can I find from the ICP. Clay is a third-party tool that Cowork can help you with: once you have the list it can upload it to Clay. You need a separate Clay subscription, but you can upload all the contacts as a CSV or Excel, and Clay does the enrichment. Enrichment means making sure that when you extract data from Apollo, ZoomInfo or any B2B source, the prospects have the right email address, contact details and name, because otherwise your emails might bounce, which hurts your deliverability and domain reputation.

Before you run a cycle, there is a twenty to thirty minute setup on Claude Cowork, or whichever tools you use (you can probably do very similar things with others). If you use Claude Cowork: marketing, sales and productivity plug-ins installed, at least one data connector (Apollo or ZoomInfo), at least one enrichment tool (Clay or Lusha), and a sequencer connected. You will use something like Instantly for the outreach itself, and the CRM and the rest can be connected too. Happy to share more information after the call.

Email sending tools

I will go back to the slides and show you the rest: the tools you can use, the third-party tools we have tested across multiple campaigns and sectors and believe work. For your AI revenue engine to work, Cowork, Claude or OpenAI cannot send an actual email themselves. You need to connect them to an email-sender tool, a third-party tool. Instantly is one good option; Smartlead can be good; Apollo can supply data and also send emails; and there is Reply.io, Mailshake, QuickMail and Salesloft. These send emails on your behalf. If you connect them as connectors via your OpenAI or Claude account, then everything you developed in the previous steps, all the copy, can be sent.

LinkedIn outreach

I want to make this very clear: LinkedIn is extremely cautious about which third-party tools you use on their platform. They do not like too many third-party tools and automations, and they can even ban your account. Sales Navigator is a good, safe solution: if you have it and connect it with your AI tool, that is the safest thing you can do. All the other third-party tools are tested and widely used across sectors and niches, but they come with some risk, because if LinkedIn's algorithm detects that you are using automation to send connection requests at scale, depending on the tool, it may act. Please check each tool you want to use for compatibility with LinkedIn's terms and conditions, and if you are not confident, just use Sales Navigator. Make your decision based on your own needs.

I am Peter Juhasz, Chief Executive Officer of Syrvi AI, and thank you, Ash, for the intro. Welcome everyone. This afternoon we are going to talk about the AI revenue engine. In the first part I will cover what the AI revenue engine means, its purpose and why it is so important, and how we built it for ourselves within Syrvi. In the second part I want to give you something that is not just a concept, but how you can execute it, even build it yourself if you want to. That part is more about execution, and I will share some of the secrets too, the tools we use and everything else.

As Ash rightly mentioned, SMEs, small to medium businesses, are facing a critical decision: whether to implement AI, although I would rather say when and how to implement AI. Enterprise-level companies can spend a huge amount of money to implement the kind of revenue engine I am going to show you today. But this is not just about the concept, it is about the practical framework for how you can build a predictable pipeline yourself, where the different elements of lead generation work in a cohesive manner, building your pipeline and, more importantly, compounding over time.

The sales paradox

The first slide is the reality, the sales paradox. What is happening in the market? You need predictable revenue, and for that you need a strong pipeline. But you cannot operate from the same budget as enterprise companies. They have anything from 50,000 to 200,000 available annually, sometimes even monthly, while on the SME side you might have a maximum of 10% of that budget to build such a system. So you need to achieve similar results with perhaps a maximum of 10% of the budget, which is not easy. Thankfully, in today's world, with the latest technology, it is possible.

The reason SMEs struggle is the reliance on isolated tools and isolated tactics, one of the main causes I have seen across multiple sectors and niches over the last ten years. By isolated tactics I mean: some companies use cold emailing, sending emails in big numbers until it just becomes spam; others do some LinkedIn outreach, sending connection requests, which creates some noise but no real results; and others say, "I am posting once or twice a week, or twice a month, on my LinkedIn profile, why do I not see results?" Content is great, super important and critical, but content alone will not solve the problem, it is just too slow.

From isolated tactics to a revenue engine

So how do we move from isolated tactics to what we call a revenue engine, an integrated system that compounds? The first layer is outreach, responsible for building immediate pipeline. You run cold email campaigns for a prospecting list of 500, 5,000 or 50,000 companies, depending on your sector and niche. You can also send LinkedIn connection requests: with the same ICP, the ideal client profile, you connect with the same people on LinkedIn, which adds an extra layer on top of email. But one of the biggest problems we all face is that when you reach out, most people do not know who you are. That is one of the most important reasons outreach fails on email and LinkedIn.

How we solve this is by implementing content, what we call a content flywheel. Content alone is too slow, but by content I mean posting thought-leadership-type content on your LinkedIn profile two, three, four times a week, sometimes every day. As long as it is high quality and resonates with and addresses the pain points of your ideal client profile, it accelerates your outreach. Before people respond to a cold email or LinkedIn message, they tend to check your profile and check who you are, whether you are the one who can solve their problems, and only then respond. So content helps accelerate your outreach on email and LinkedIn.

The third thing is an automated qualification and nurturing system, even a booking system. Leads can come in different ways: from outreach, where people respond to your email or LinkedIn, but prospects can also come to you via your website or your LinkedIn content, filling in forms on your website, however they reach you. You want a qualification system that qualifies them and books them with you, or with your sales team, into the calendar. After qualification and booking, you have the call. This is the momentum we build with the revenue engine, a flywheel that accelerates over time, running 24/7 and creating that momentum.

Pillar one, intelligent outreach

The first of the three pillars is intelligent outreach. You need a highly targeted prospecting list, you need data to generate immediate pipeline. It is not just about sending emails, it is about sending the right emails to the right people with the right messaging. We start with data and smart segmentation. Data always comes first, then segmentation: by sector, niche and subniche, going down three or four levels, because every industry has its subniches and different companies specialise in different things. You also want to segment by role: within a company, is it the C-suite, director level, manager level, or below? Who do you want to target? This is why segmentation is so important, you want your outreach landing, via email and LinkedIn, with the right message to the right people, orchestrated across a seven or eight step sequence, and personalised.

With the right data and multi-level segmentation, the AI agents can write the right content for the right prospects. Your message does not come across like a templated outreach message that everyone else sends, it comes across as someone who understands their business and their pain points. Then the final step in this first pillar, outreach, is the execution itself: using email-sender tools to send the emails, and sending very personalised messages on LinkedIn. Importantly, all of this works in parallel.

The intelligence layer, a knowledge base

Before we go to the content flywheel, you might ask: how can I make sure the AI agents do not make things up, or send something out of context, or worse, things we have all heard stories about? One of the most important things we built into our AI revenue engine is what we call a knowledge base, a vector database. I will not spend too much time on this, it can get very technical and that is not the purpose, but it is important to understand that a knowledge base for AI agents is very similar to one for humans, like a super-intelligent SharePoint where we share information and always have access to the right context.

For any AI agent to write copy, like any human, it needs context: about your company, your brand guidelines, your values, your value proposition, and information relevant to your ideal client profile, who the client profiles are and all the segmentation. All of that is critical for AI agents to write the right copy, because if they do not have the right access, or have the wrong information, that is where hallucinations start. You can build this knowledge base yourself, or with someone else. The way we built it includes the knowledge base not just for the AI agents but also for the humans: any work we do internally, within a team or with the AI agents, all uses the same source of truth. That is really important, because if teams have different or wrong information, whether human or AI, they will make things up, which we do not want.

Pillar two, the content flywheel

I already mentioned the importance of the content flywheel. Its first purpose is to create good-quality content for your LinkedIn profile, or even your website if you want to do high-quality content like blogs. Here I will focus on LinkedIn. LinkedIn matters because you want to position yourself as an expert in whichever segment and ideal client profile you are targeting, as someone trustworthy who can solve those pain points. That builds your credibility, because when you send emails and connection requests, people check your profile before responding. This is a long-term asset, a mid to long-term game. I always say: do not expect people to start coming to your posts just because you start posting every day, and do not think you will generate leads from day one. What it does is accelerate your outreach: more people respond, and in the mid to long term, after two or three months, the way the LinkedIn algorithm works, if you know your ideal profile and consistently send connection requests to those people while developing high-quality content that addresses their pain points, the LinkedIn algorithm starts to become your best friend and shows your content not just to your connections but to anyone else who might be interested. Everyone matching your ICP, those companies and roles, will also see your content, and that is where inbound lead generation can start from LinkedIn: people engage with your post, comment, and even send you connection requests asking for your services and meetings.

Pillar three, automated qualification and booking

The third pillar is what I call the automatic qualification system, which can also be part of the nurturing system. It is not just about people responding, but also about those who somehow disappear after a first positive response. That is where we follow up, to make sure they get extra content, because some people need more time before they commit to a call. Whether or not we use voice-agent technology: in today's world people are still very careful with AI voice agents, and I understand why. We use it, and have implemented it across a couple of clients, but certain sectors and businesses are cautious about it, which I fully understand. The time is here, and it will become more and more common. By that I mean, for inbound calls, instead of a human, an AI voice agent picks up the phone with a set of pre-qualifying questions: is this an unsolicited sales call, or someone relevant to your product and service? The AI agent goes through those questions and books relevant prospects into your sales team's calendar, because that is where you should be spending your time, whether you have a sales team or you do the selling and closing yourself. You should not spend your time prospecting and on outreach, nor qualifying people, because that can be done by an automated nurturing and qualification system. You should spend your time closing, with prospects who have already been qualified and booked, so they are highly relevant, not just anybody.

So the whole flywheel effect is created by the three pillars: one, outreach on email and LinkedIn; two, content for your LinkedIn that increases your touch points and accelerates your immediate pipeline from outreach; and three, with or without voice agents, an automated qualification and booking system. It does not need a voice part, it can just be SDR-type agents, sales development representative AI agents, that ask the right questions via email, chat, WhatsApp, Telegram, or whichever platform you use. Voice-agent technology is just on top, if you need it, but it is not required for the qualification system to work.

The go-to-market growth plan

So how would you do this, and what tools would you need? I will stop sharing for a second and move to the second part of the presentation, which is more about showing you how you can execute it yourself. Before you execute a plan and build an AI revenue engine, one of the most important things you need is the plan, what we call a GTM, go-to-market, growth plan. With every client we work with, we always develop this plan. It is critical to understand that, depending on your sector, niche, service and offering, you want to know what you are doing. Without a plan and strategy, it is very hard for any revenue engine or AI system to execute properly, because this integrated system, these three pillars, work simultaneously.

Typically it looks like this. There is an executive summary about you as a company, your offerings, and the TAM, the total addressable market. There is how the outreach will work and what you can expect, and there are intent and buying signals: how we can find them for your business. There are common intent signals like growth and expansion signals, hiring signals for key roles that indicate the type of service they need, leadership and strategic changes within companies, and so on. In different sectors and segments we tap into different intent signals, which helps get the messages out not just in a personalised manner but a timely one, because timing is critical. If someone is searching for the type of solution you offer, they are more inclined to reach out or respond.

Service alignment is important, as is the total addressable market. We break down what kind of service you offer, then break down every segment, showing exactly how many companies we can find within each, the number of decision makers, and the priority: high, medium-to-high, medium. We always determine at least four or five different segments and campaigns.

Tiered targeting by role

This is super critical, and very few people talk about decision makers by role category. Why does it matter? What we have found running outreach campaigns for over five years is that the size of the company makes a world of difference to which title and person you target. For example, say we have a value proposition relevant to ESG and sustainability offerings. It could be that we do not target the founders, directors and owners that most people would default to. Yes, target founders and owners first, but the problem is that if the company is more than, say, 10, 20 or 30 employees, founders, owners, even C-suite, will not answer, or will answer a straight no. But if you target relevant titles, in this example head of social value, social value manager, ESG director, ESG lead, sustainability director, even supply chain director, you are more likely to get a response. So we normally do tier one and tier two targeting within the ideal client profile. Tier one are the people most likely to respond to our outreach; tier two are who they will escalate it to. A CEO of a 50 or 100 employee company might not respond, or will say no, but a head of social value or social value manager will escalate it to the CEO, because the pain point is different. The pain point of the CEO or founder can be completely different from certain other roles. This is why the segmentation I showed you is so important: if you do not get the segmentation and the roles right, you start sending messages tackling the wrong pain points.

ICP breakdown, playbooks and the ICP of the ICP

We also do an ICP breakdown. ICP stands for ideal client profile, and we break it down by campaign. Campaign one, for example, construction contractors: we detail the industries, locations and revenue, the size of companies and employee count, and every company has a tier one and tier two job title for targeting, plus company keywords. Then campaign two, the same detailed breakdown, and campaigns three, four and five the same.

What I am not going to show you today is that for every single campaign in this go-to-market growth plan, we also develop a playbook. A playbook includes every role within tier one and tier two that we are targeting, say CEO, manager, director, chief commercial officer, head of product, every title gets messaging and pain points depending on the size of the company, the angles we will use, what angle we will tackle, a pain we will address, or something we will offer, and the status quo, what would happen to them if they did nothing. It is a very detailed playbook for every person we target within each company.

After the go-to-market plan, we also develop the ICP of the ICP, the ideal client profile of your ideal client profile. Very rarely do I ever hear about the importance of this. It is so critical: for your message to land and not come across as a templated outreach message that anybody can send on LinkedIn or email, you need to know who your ideal profile is selling to. That gives the AI revenue engine, the AI agents writing the copy, the information to make it land, so it comes across as, "this person and this company understands my problem and my pain point, and is going to solve it." It is about landing your message the right way.

Proposition, messaging and roadmap

Once we have all this information, we develop a proposition by segment. Depending on your offering and services, we break it down by segment, so different sectors and niches get different messages, with your key differentiators, your USP, your unique selling point, why you are different and why they should choose you over others. Then the example positioning and angles come for the cold outreach. Every campaign and every tier receives a different message: LinkedIn connection requests, email subject lines with multiple variants, email opening lines, and the follow-up sequence, how email and LinkedIn work together, all based on the previous slides. Without that information, the AI agents in the revenue engine would not know who to send to, or what email copy and subject lines to write, and you want to stand out when you land in someone's inbox. So every campaign has different messaging: a LinkedIn connection request relevant to and referencing the particular sector, niche and pain points; subject lines relevant to that sector and niche; and opening lines relevant to the segments we are targeting. Everything is personalised.

We normally do an implementation roadmap for the first 12 weeks: what the outcomes look like after 12 weeks, the success metrics, and what they can get out of the plan. Now this is where I really wanted to give you something you can use and take away immediately: how I would build an AI revenue engine. So let's say we do not have this system within Syrvi, we have no software. What would I...

Myles: It is a pleasure. I am really excited about this topic, because this is what every business owner is curious about: how AI and humans can work side by side to help SMEs grow faster, smarter and stronger. Peter, let's kick things off with you. After our last episode, one thing that stood out to me about Syrvi AI is how you do not frame it as AI replacing people, but rather as AI joining the team. Most companies either go all in on automation, or keep things completely people-led. Why do you think this hybrid model, AI plus human, is the real sweet spot for SMEs?

Peter: Thank you, great question. I think there is still a misconception about what AI can and cannot do, and that causes fear among people, employees, and sometimes business owners too. We experience this as well. The hybrid model works because it leverages the irreplaceable strengths of both humans and AI. We all have strengths and weaknesses, and so does AI. Pure automation does not work on its own, because business is ultimately about relationships, and relationships require empathy, creativity and the nuanced understanding that only humans possess. Conversely, purely human-led sales teams are limited by time, energy and consistency. There are things humans excel at, and things AI excels at. Our hybrid approach lets AI handle the heavy lifting, qualifying thousands of leads at scale, personalising tens of thousands of leads, creating content and managing follow-ups, so human teams can focus on what they are really good at: emotional intelligence, closing deals, building relationships, creativity. That is the message we always pass to our clients: let your team, and yourself, grow using the skills you have, and let AI do the rest and empower your team.

Myles: That makes so much sense, Peter, because at the end of the day business is still about relationships, and you cannot automate empathy. I like how you frame it as AI amplifying the human side rather than replacing it. So, Istvan, building on that: once AI takes over the repetitive, time-consuming, or boring tasks, what door does that open for human teams? Where do you see leaders being able to focus their people power?

Istvan: Yes, very good question, and one I get asked a lot. We look at this as a bit of a liberation process. When AI liberates your team from those repetitive tasks, a few things happen. Firstly, strategic thinking: your team can finally analyse market trends, develop new offerings and innovate, rather than just execute. Secondly, relationship deepening, working on the relationships we humans really do with one another. Instead of constantly chasing new leads, they can nurture existing clients, which increases the lifetime value of those clients, which is really important for the business. The third opportunity is skill development: once freed from the mundane, repetitive tasks, your teams can upskill in areas like consultative selling, industry expertise, and even strategic account management, which is a key aspect of growth. I have boiled down where leaders should focus their people power into three Cs: creative problem solving, complex negotiations, and client success. These are the areas where human intuition and emotional intelligence create exponential value that AI, at this point, cannot replicate. We do not know what will happen in the future, but at this point that is where the human element is really powerful, and that is where the combination comes in really well.

Myles: So the powerful three Cs: creative problem solving, complex negotiations, and client success.

Istvan: Exactly.

Myles: I love that. Waking up and finding your calendar already filled with qualified appointments while AI is quietly working in the background, that is more than just being efficient, it gives peace of mind. Of course, with every new technology there is the human factor. So, Istvan, in your experience, how do employees usually react when AI suddenly joins the team? Is it excitement, resistance, or somewhere in between, and how do you help companies manage that transition?

Istvan: In our experience the reaction is surprisingly positive, but it has to be positioned correctly. Most employees are actually frustrated by repetitive tasks. I do not know many people who really enjoy the mundane things; people enjoy it when they can truly add value and do meaningful work. That means different things to different people, but repetitive, tick-box exercises, doing things over and over, most people do not enjoy that. So when people understand that AI handles those mundane tasks so they can focus on more strategic things, the excitement starts to build. We see three phases. First, initial apprehension, because not everyone likes change, that is a fact. Then a curious experimentation phase, where people think, "I will just give this a try and see what comes back." And once that is over, which happens reasonably quickly, they move to advocacy: "this actually works, this is something I am happy to continue with." There is a transformation when they experience that first AI win, which is really useful, because that is how we humans learn, through positive reinforcement. If something works and helps me, I start to like it; if it makes my life more difficult, of course I will not. But we tend to have the positive aspect: if someone can close a deal they would not have otherwise, that is a great win; if they can finish work on time, or even earlier, because AI handles things that normally get pushed down the priority order but still need doing, often at 5pm when you are meant to be going home to pick the kids up, those are really good wins. We help companies navigate this by creating AI champions, people who get behind it and understand it from the get-go and celebrate the collaborative wins. The data speaks for itself: an overwhelming number of employees report higher job satisfaction than before they started using it. Overall, the picture is quite positive.

Myles: That is a very refreshing take, Istvan. Instead of fearing replacement, people are finding more satisfaction in their work by adding value to the organisation, once AI takes over the boring, repetitive stuff. A lot of us will find comfort in hearing that. So let's look ahead. If we fast forward five years, what role do you see this human-AI partnership playing in the way SMEs compete with much larger players?

Istvan: It is an interesting one. We have a lot of conversations about this: where will this be in two years, in five years? It is always difficult, because the speed of change with AI is rapid. But our take is that in five years, human and AI synergy will not be an advantage, it will be the minimum requirement for an SME's survival. That is the bottom line. We are seeing a lot of uptake from small to medium businesses, so the ones that do not adopt will find it tricky quite quickly and will have to start catching up. The exciting part is that I think SMEs will actually have an edge over enterprises, which is strange, because normally it is the other way around. Large corporations struggle with legacy systems, and change in larger organisations is harder to manage. As previous companies I worked for always said, turning an oil tanker around is a lot more difficult than turning a little pedal boat. An SME is like a small boat you can row, versus a large enterprise that is more like an oil tanker. So we are going to see David-beating-Goliath stories, where, for example, a 10-person company starts winning against a 1,000-person enterprise, because they were able to adopt and integrate these AI technologies. Some SMEs will master these synergies and operate with the efficiency of algorithms but the creativity and adaptability of a startup. We will start to see markets that were previously impossible due to resource constraints open up. I predict we will see some billion-pound or billion-dollar SMEs emerging over the next five years. It is hard to predict, but small businesses can adopt very sophisticated AI systems and compete globally while maintaining their local authenticity, which is what makes them SMEs. Interesting times ahead. That is how I see it. Would you like to add something, Peter?

Peter: Very well summarised. Honestly, it is hard to predict; I do not even want to look five years ahead. I think even one or one and a half years will bring significant changes, I do not think we will recognise in five years what is happening now. We see the speed of AI development and improvement, and five years is just too far away, so much will change. Complex sectors and industries will be completely changed, or even disappear, and some jobs will disappear, but other jobs will be created because of this. There is always an element of change, but in five years I see huge and very significant changes, in a good way, for SMEs, which will be in a better position. One more thing I think is important to mention: I sometimes hear that the AI bubble is going to burst. It is not a bubble. AI is here to stay, and it is only going to get stronger. It is like when computers came in the 50s and 60s, then Microsoft, then Excel. At first people did not even know what to do with a computer; no businesses got it. Then Excel and the rest of Microsoft came out, and now we cannot even survive without computers. AI is going to be even more significant than the computer was in the 50s and 60s. It is never going to turn back; there might be slight dips, but AI is here, and it is going to change our business and our personal lives forever.

Peter (host): Hello, hi everyone. Thank you very much for coming today to talk about this amazing subject: AI, and particularly how AI can help small to medium businesses with operational efficiencies, sales and lead generation, and the challenges we as SME owners face. It is an honour to share the panel with so many great people, and with you who came today. Let me introduce the panel: Chris Burgess, Mariam, Steve Hilton and Viv. The first question we will discuss is how AI is reshaping the SME industry: what operational efficiencies we can achieve, and what challenges we face as small business owners.

[The panel discusses coming back to business basics and chasing an ROI rather than a trend, AI as a tool that shifts work rather than replacing people, and the importance of an AI policy and data security.]

Peter: I could not agree more with everything my fellow panellists said. AI is a great tool, but you need to know how to use it. For example, why would you prospect people manually on LinkedIn, or search databases, when AI can do that for you, and many other things? But you need to make sure it is safe: have an AI policy, self-hosted if possible, or even more secure, so you do not share your client data with anybody, and get approval from your clients on how to use AI if you use it for them. If you use it for yourself, that is your decision. Now the third question: what does the future hold for us humans working with AI? How can we complement and empower each other, and use AI in the right way, so it empowers us, rather than thinking, "they are going to take my job, so I will not bother"?

[The panel shares views on AI making life easier, freeing people for higher-value and creative work, and interacting with AI by voice in future.]

Peter: It is a great subject, and I often wonder where we are going with this too. A lot of great minds will develop amazing products and better AI, but it should all be based on what we need, the way we interact with AI, and how they learn from us. It is about collaboration: letting AI learn from us so it gets better and takes over the things we should not be dealing with. My take is that we humans should focus on two things. One, our creativity: I do not think AI is creative in the way we humans are, so we come up with the great ideas and AI helps us execute them. Two, AI can help us become better.

[An audience member asks how AI can support a startup; the panel recommends starting with a freely available tool and a clear strategy.]

Peter: We also do a full, free business transformation report for everyone during the expo: where you are, your niche, your industry, your sector, who your competitors are, where they are beating you, and what gaps you can exploit. AI is perfect for that, it can go through hundreds of thousands of data points that we cannot as humans. Strategy is critical, and to have a strategy you need an audit of your competitors and your market.

[Closing the panel.]

Peter: My closing advice: test multiple models, and test the prompting. Prompting is very important, the system prompt, the user prompt, and the context is critical, you need to give the AI exactly what you want so it gives you the right answer. Test it multiple times, on the same model or across multiple models, compare the answers, and analyse which approach is best. If you want new customers, you need a good lead generation strategy: you need to know who your customer is, what niche they are in, what pain points they have, and then, with the help of AI, develop the right message and the right channel to deliver it. That is the only way I can imagine to generate leads and sales for your business today. I hope it helps, and you can come and see us at Syrvi for a free strategy. Thank you to Chris, Mariam, Steve and Viv, and most importantly to you, who turned up.