Notes
Replicating Gartner’s 67% benchmark
Gartner said 67% of B2B software buyers prefer a rep-free experience. When we asked our panel naively, it failed completely. When we added real sales friction, it revealed what actually happens.
If an AI panel just gives you polite corporate answers, it is worse than useless. We tried to replicate a famous industry survey. The mistake taught us more than the success.
What Gartner reported
In March 2026, Gartner published a major survey of business software buyers. Two numbers caught everyone’s attention:
- 67% of buyers said they prefer a rep-free, self-serve buying experience.
- 45% of buyers said they use AI tools (like ChatGPT or Copilot) during procurement.
We set up a Synthser panel of 100 enterprise decision-makers — Managers, Directors, and VPs across IT, Finance, Operations, and Marketing evaluating $15k to $50k software — to see if our synthetic panel would reproduce the same numbers.
First try: The polite AI trap
The first time through, we asked an abstract question: "Do you prefer working with a sales rep or a rep-free experience?"
The result was a total failure: 92 out of 100 personas chose the sales rep, and only 7% chose rep-free. AI adoption came in at just 12%.
AI models are trained to be polite and corporate. When an AI roleplays a "responsible VP of Finance", it defaults to textbook behavior: "A good executive meets with people, reviews contracts together, and builds partnerships."In the real world, human buyers hate sitting through 30-minute qualification calls just to see a pricing page.
Second try: Adding real friction
Real buyers do not hate sales reps out of philosophy. They hate lost time and hidden pricing. We rewrote the question to describe what actually happens in both buying paths:
Choice A
Self-Serve
Instant 14-day interactive trial, pricing published upfront on the website, and security docs you can read on your own time without scheduling a meeting.
Choice B
Sales-Guided
Submit a demo form and complete a mandatory 30-minute discovery call with an account executive before getting pricing or sandbox access.
Once the choices included real-world friction, the panel woke up:
- AI tool adoption jumped to 38% (landing right inside Gartner’s 35%–55% range).
- Buying preference split exactly 50 / 50.
The results side by side
| What we tested | Gartner survey | When asked naively | With real friction |
|---|---|---|---|
| Rep-free buying preference | 67% prefer self-serve | 7% (Polite AI default) | 50% (The real split) |
| AI tool use during buying | 45% use AI tools | 12% (Too conservative) | 38% (Replicated) |
| Buyer confidence | Mixed / hesitant | Clustered together | Clear split by company size |
| The takeaway | Broad headline | Do not trust naive prompts | Build self-serve, keep reps for legal |
What the 50/50 split actually means
Why did our panel stop at 50/50 instead of 67%? Because at $15,000 to $50,000, software sits in a unique "procurement dead zone":
- Mid-size companies (50–500 people): Strongly want self-serve. They move quickly and do not have complicated procurement layers.
- Large companies (500–5,000 people): Love the idea of self-serve, but their legal and security teams require human negotiation before any contract gets signed.
At this price point, software is too expensive for an unapproved credit card swipe, but too small to justify a painful three-month enterprise sales cycle.
If a founder read Gartner’s 67% headline, they might fire their sales reps and try to go 100% self-serve. That would fail the moment enterprise deals hit legal. The smarter play is a hybrid model: give buyers instant self-serve access and public pricing, but keep a human ready the second procurement and security show up.
Show the work
Synthetic market research is only useful when you show the mistakes alongside the wins. If we hid our first attempt, you wouldn’t see how polite AI models can mislead you. Every Synthser brief shows the whole setup so you can decide what to trust.