Most B2B companies responding to AI are doing it by buying things: a new writing assistant, a new AI sales tool, an AI-powered analytics layer. Sometimes all three in the same quarter, from three different vendors.
The instinct is understandable. AI is changing how marketing works, and buying a few tools might feel like responding to that change. But most AI tools in marketing don’t automatically create new capabilities. If your attribution is clean and your lead definitions are consistent, they save time and sharpen your decisions. If they aren’t, they produce wrong answers faster, at lower cost, and with a lot of confidence.
If you’re running a SaaS or lead generation business between €1M and €10M in revenue, read further to understand what you can do.
Measuring what’s working has become harder
The first huge shift is in measurement. Although it’s been building for a few years, AI has made it way more obvious.
Most B2B marketing at the €1M to €10M revenue stage is built on a measurement model that no longer works as well as it used to:
- Third-party cookies are effectively gone
- LinkedIn controls what your content reaches and on what schedule.
- Google and Meta have progressively removed the granular data that used to let you see which campaigns were doing the work.
- And LLM search (ChatGPT, Perplexity, Google AI Overviews) – is quite a black box. You don’t see who searched, you don’t see what they were told, and you don’t know whether you appeared until a prospect mentions it on a call with you.
The practical consequence: most of the companies we speak to are still running campaigns and reviewing dashboards that paint a confident picture of the buying journey. That picture is increasingly reconstructed by the platform’s own algorithm, not by what happened.
A more reliable approach is to build measurement around 3 things together:
- what the platforms report (imperfect, but directional),
- what prospects tell you at the moment they convert – “where did you first hear about us?”
- and whether the people you attracted in a given quarter turned into revenue
Any one of these on its own gives you a partial picture. The combination tends to tell the truth.
It’s unglamorous work. But without it the rest of your measurement is guesswork. And AI tools layered on top of that guesswork produce confident-looking guesswork.
Your data quality is amplified now – in both directions
The second shift is the one most companies underestimate.
Every single AI tool you add to your marketing runs on the data you already have:
- AI lead scoring trained on incomplete CRM data produces wrong scores faster
- AI attribution fed messy conversion tracking produces wrong attribution at scale
- AI personalisation built on partial customer data produces generic output that looks personalised. But isn’t.
The companies that invested in clean data infrastructure through 2023 and 2024 (like accurate tracking, a shared definition of what counts as a qualified lead, consistent CRM records) can now add AI tools and trust what they produce.
The companies that skipped that work are in the opposite position. The mess they had before is now a significantly larger, overwhelming, faster mess.
The most valuable AI investment you can make is often not an AI tool at all. We say this to almost every founder we speak to. It’s the focused work that fixes what’s underneath:
- attribution setup
- CRM hygiene
- conversion tracking
- ICP definition
Do that, and the tools you add afterwards have something reliable to work with.
What the marketing team needs to focus on
The third shift is internal, and it tends to show up rather slowly.
Marketing’s job has always been to understand the market and translate that understanding into pipeline. What AI changes is the balance of what takes time. The routine parts of production (first drafts, report formatting, basic research) are increasingly automatable. The parts that aren’t are the judgement calls:
- what to say to which audience at what moment
- how to frame a problem your prospect hasn’t yet named
- what the data is telling you versus what the dashboard is designed to show
This changes what a good marketing hire looks like, and it changes what a good agency relationship looks like.
The question to ask about both is the same: do they take accountability for what happens to the business (qualified leads, pipeline, closed revenue) or do they take accountability for the activity that’s supposed to produce those things?
In a market where AI has made campaign production faster and cheaper, the deliverables layer (the ads, the posts, the reports) is the easier part. What’s harder to replicate is the thinking behind the activity, and the willingness to own whether it moved the business.
The pricing model matters more than most companies realise
This is the part that rarely comes up until after someone has been burned. Standard retainer pricing creates a specific dynamic: the agency is paid regardless of what happens to the client’s business.
There’s no structural reason to flag that attribution is broken, or that the lead definition doesn’t match what sales is trying to close, or that the campaigns are optimised for clicks rather than revenue. The monthly report looks fine either way.
The client often doesn’t realise until months and budget have passed. It is one of the most common stories we hear from companies that come to us after a previous agency.
Performance-based pricing changes that dynamic. When the agency’s fee is tied to qualified leads or pipeline generated, the agency’s incentive and the client’s incentive are the same.
There’s now a structural reason to fix the attribution before running the first campaign, to agree on what a qualified lead actually means, and to say clearly when a client’s foundation isn’t ready to support paid acquisition.
In a year when AI has made it easier than ever to generate activity, the pricing model is one of the clearest signals that determines whether a click becomes a lead, and whether a lead becomes a closed deal.
It’s worth asking about before you sign anything.
Additionally, you also should look into one-month notice contracts. If a hired agency is not delivering, you should be able to leave next month.
Where should you start
The instinct when AI arrives is to ask what to deploy. The more useful question is what your marketing needs to look like in two or three years to still be working, and what you do first to get there.
For most SaaS and lead generation businesses, the answer tends to follow the same sequence:
- Fix the data infrastructure first: attribution, CRM, conversion tracking and a shared definition of what a qualified lead is. Without that, most of what follows is difficult to evaluate honestly.
- Then invest in the marketing activity (paid acquisition, content, channels) that your now-sound infrastructure can evaluate and act on.
- Then look honestly at whether your agency relationships are structured around accountability for pipeline, or accountability for delivery.
Done in that order, the AI tools you add along the way have something reliable to work with, and the results show up in your pipeline. Not just in your dashboard. Done in reverse – tools first, structure later – you’re optimising what you can’t properly measure.
If you want to know where your foundation stands before you change anything, our pipeline scan does that in seven days. Free, no obligations. We tell you what we would fix before we would touch your spend, even if you do not end up working with us.
FAQ
What’s the most important first investment for a B2B company at €2M revenue?
Fix attribution before anything else. The specific question to answer is: can you currently trace your ad spend to closed revenue? Not form fills, not demos booked – revenue. If not, that’s the work to do first. It typically takes several weeks done properly, and it makes every subsequent marketing decision more reliable.
How do I know if my data infrastructure is the problem?
Three questions: do your sales and marketing teams agree on what a qualified lead looks like? Does your CRM accurately reflect where leads came from? Can you connect a specific campaign to pipeline generated, not just leads? If any of these is no or uncertain, the infrastructure is the problem, not the channel.
Does traditional SEO still matter when more prospects are using LLM search?
Yes. There is growing evidence that content performing well in traditional search also appears in LLM answers – specificity, authority and clarity matter in both environments. What’s become far less effective is generic content produced purely for volume. Specific, opinionated content based on real data or genuine experience tends to hold up better in both.
Is performance-based pricing right for every company?
No. It requires sufficient lead volume to make results measurable, a clear definition of the outcome being tracked and infrastructure that can verify what happened. Companies whose tracking and CRM aren’t yet in order are usually better served by fixing that foundation first — then moving to a performance model once results are properly attributable.
