It’s been a big couple of weeks for investment into AI consulting services (you can read our deep dive on OpenAI and Anthropic strategies here). If you spend any time with founders of AI Enablement companies, you hear the same paradox again and again: demand has never been higher, yet turning that demand into a clearly scoped SOW is extraordinarily hard. In fact, many think it is more difficult than closing almost any deal in their prior careers.

To understand why, we surveyed 100 senior decision-makers in mid-market and enterprise companies who have actually purchased AI enablement services (meaning consulting, training and/or implementation) in the last 12 months: budget owners, final decision-makers, and procurement leads. Most market analysis in AI enablement is written from the supply side: what providers are building, what investors are funding, what founders are pitching. The buyer's voice has mostly been silent.

89% say purchasing AI Enablement is more complex than buying any other professional service. 76% say slightly more complex, 13% say much more complex. Zero found it easier. These are experienced buyers of management consulting, IT services, L&D, and legal services. They are not new to procurement. And they are telling us this emerging space is significantly harder.

Every AI Enablement provider should let that number inform how they think about go-to-market.

The survey breaks down exactly where friction is coming from. Security and compliance concerns top the list at 31%, followed by the rapidly changing AI landscape at 22%, and difficulty comparing vendors on like-for-like criteria at 19%. Cross-functional decision-making – HR, IT, and business units all involved – adds another 12%.

Read these not as buyer complaints but as intelligence about what your sales process is running into.

Security and compliance at 31% means buyers are arriving at vendor conversations without internal governance settled. Your first conversation is often about their unresolved policy questions, not your capabilities.

The rapidly changing landscape at 22% means buyers are suffering from evaluation paralysis. They've seen enough vendor pitches to know that what's true today may not be true in six months. A defensible decision is harder to make when the ground keeps shifting.

Difficulty comparing vendors at 19% means your positioning probably looks like everyone else's to the person trying to justify a decision internally. That's a structural problem, not a messaging problem.

But not every type of service runs into the same friction.

Buying AI training is not fundamentally different from buying other training services, except that most buyers struggle with two specific things: internal ROI discussions, and picking the right training format. Should one vendor train the whole company? Is a function-specific approach more likely to get engagement? In the absence of broader AI initiatives, how much can training alone move the needle?

Despite all this, I'd argue training remains the easiest entry point into organisations. But without a deliberate ROI conversation and real “training-format-to-customer-fit” thinking, it's unlikely to generate recurring engagements. Which probably makes it a difficult business to be in over the medium and long term.

AI consulting and custom workflow implementation are harder to sell upfront because the buyer has to expose their internal operating model and sensitive data. This is where the security concerns in the data come alive. Helping customers resolve their internal governance isn't strictly your responsibility as a vendor. But having someone in-house who can help alleviate these concerns (a fractional CISO, a compliance-fluent advisor, anyone who can accelerate the buyer's internal process) is a growth catalyst. It often makes the difference between a stalled conversation and a signed contract.

European buyers are operating under a stricter regulatory environment, a less mature local vendor landscape, and higher internal scrutiny on AI-related procurement decisions. The cost of getting the selection wrong is higher, which makes the decision harder to make confidently. And one of our survey findings illustrates this tension clearly.

For any founder building across markets, or investors assessing European AI enablement businesses: this friction is real, and it is not resolving quickly. The implication isn't that Europe is a harder market. But the playbook for winning there needs to account for a buyer who is more risk-averse and less able to evaluate independently.

Various government schemes and subsidies are active across European markets to help companies stay competitive. This is most visible in training – we covered it in our Multiverse deep-dive. Navigating compliance and country-specific requirements naturally concentrates operators on single-country markets. And because government incentives are widespread, you can build a solid business in your home market without crossing a border.

But that's only part of the explanation and it doesn't account for why categories like custom AI apps for internal use cases are genuinely harder to sell in Europe than in the US.

One hypothesis: the breadth of founder backgrounds is much greater here. In the US, the majority of AI enablement founders come from the venture-funded world; either as repeat founders or early employees at VC-backed companies. In Europe, we meet all kinds of founders: former execs of large established companies, founders of traditional consulting businesses, younger founders who started essentially straight out of university, engineers and product people making the jump to services. Despite talking to 100+ founders over the past six months, we have yet to meet two companies with exactly the same approach and product offering.

That diversity is genuine and, I think, an underrated strength of the European market. But it also makes vendor comparison harder for buyers – which probably explains a meaningful share of the 19% who struggle to compare vendors on like-for-like criteria. When every provider has a different background, a different framing, and a different pricing structure, there's no common vocabulary to compare against. Buyers aren't evaluating like-for-like proposals. They're trying to translate between fundamentally different offers, which is exhausting and often ends in a decision based on whoever felt most familiar, not most capable. That same dynamic is almost certainly part of why only 19% of buyers feel confident evaluating vendors independently, and why 57% fall back on existing relationships to source options in the first place.

Here's the critical insight from the data: complexity doesn't make buyers more rigorous. It makes them more conservative.

When evaluation criteria break down, buyers fall back on trust. They typically evaluate just three vendors before deciding. They source options through existing relationships (57%) and peer recommendations (28%). Only 19% feel confident evaluating vendors independently. And 89% plan to expand with their existing AI enablement vendor over the next 12 months rather than run a new search.

If you're not among the three vendors a buyer considers, your capabilities are irrelevant. And getting into that consideration set runs almost entirely through relationships and reputation,  not inbound marketing, not cold outreach or feature comparisons.

The implication for your go-to-market is straightforward, if not easy: hold on to a customer for dear life once you've won an engagement. A successful delivery creates referrals. But referrals alone aren't enough, you also need to expand your product offering to create genuine upsell opportunities, because the 89% planning to expand with existing vendors is only valuable if there's somewhere for that expansion to go.

Over the past few weeks, I've become a genuine convert to Claude's “plan” mode: building out what I need and what success looks like before diving into a task. Looking back at my experience working with customers, I didn't spend nearly enough time in that “planning” mode, even for much simpler SaaS onboarding flows. I think a lot of AI enablement founders are repeating this mistake at the engagement level: moving quickly into delivery without spending enough time scoping what a successful engagement actually looks like for the buyer. Slow down before you start, not after you're stuck. The 89% expansion number will only materialise if the first engagement was set up to succeed.

Every professional services category goes through a period where buyers can't compare providers on objective criteria and default to brand, relationships, and reputation. Management consulting did it. Cybersecurity services did it. Executive search did it. AI enablement is in that period now.

Standards will emerge. Benchmarks will accumulate. ROI comparisons will become more tractable. The complexity will reduce.

But that's a 3–5 year story.

In the meantime, the market rewards companies who help buyers make the decision to choose them. Not by simplifying AI enablement into something it isn't, but by reducing the specific friction points the buyer data identifies: clearer scoping, faster pilots, more transparent evidence of outcomes. The firms that build their positioning and sales motion around the buyer's actual experience of complexity are the ones who will be well-placed when the category consolidates.

We don't have a crystal ball, but we have a few predictions for how this plays out:

The firms that will win when this category consolidates – and it will consolidate – are the ones treating that complexity as the operating environment, not a temporary obstacle. Clearer scoping, faster pilots, and earned relationships are the durable advantages. Everything else is noise.

If you haven't already downloaded the full survey, it's available for free to all subscribers here.

– Daria