Many enterprise AI conversations start in the wrong place. They start with a pitch: a platform, its features, a case study, a price and a projected return. The buyer listens, asks sensible questions, requests a proposal, and the process slows. Each passing week makes deployment less likely.
The technology is rarely the problem. The conversation is. It assumes the provider already knows the buyer's problem, when the real problem is usually invisible from outside.
Buyers no longer want to be pitched
Gartner found in 2026 that 67% of business buyers prefer to buy without a sales representative. They research independently and arrive with views already formed. A pitch asks them to accept someone else's description of their situation. Most would rather explain it themselves.
The research on what works in complex sales has pointed this way for decades. Neil Rackham's study of more than 35,000 sales calls, published as SPIN Selling, found that in large, considered purchases, the most successful salespeople talked less and asked more. They moved from questions about the buyer's situation, to the problems within it, to the implications of those problems, and only then to the value of solving them. Presenting a solution early, before the buyer had described the problem in their own terms, made a sale less likely.
The real problem is invisible from outside
In most organisations, the documented process is not the real process. The org chart does not show how decisions are made. The data that exists on paper is not the data people use. The knowledge that keeps the operation running, the exceptions and workarounds experienced people handle without thinking, cannot be seen by anyone who has not spent time inside.
Providers also suffer from what psychologists call the curse of knowledge. Camerer, Loewenstein and Weber showed that people who know something find it hard to imagine not knowing it. A provider that has seen many similar operations assumes it recognises this one. It often does not, because what makes each operation different sits in exactly the details an outsider cannot see.
A pitch cannot surface any of this. It matches outside assumptions to an outside view of the organisation. The result is a proposal for the documented problem, a pilot that works in controlled conditions, and a deployment that stalls when it meets how the organisation really runs.
Understand the job before the solution
Christensen's jobs-to-be-done theory offers a helpful frame. Customers do not buy products; they hire them to make progress in a particular situation. The question is not "do you need AI for quoting?" but "what are you trying to get done, and what is getting in the way?" The answer might be winning more enquiries, freeing two senior engineers before one retires, or passing a customer audit without a scramble. Each points to a different first step.
Starting from shared patterns
A better discovery conversation reverses the flow of information. Instead of diagnosing the organisation, it offers three or four patterns that are common at its stage, described neutrally and without linking them to any solution. For example: knowledge that lives in a few people rather than in systems; decisions made with partial context because the full picture is spread across teams; automation in pockets, with nothing connecting it.
Then it asks which feels most familiar. The patterns are honest observations from many organisations. They are not a script, and they carry no judgement.
Why this works, and why it is respectful
Research in communication and psychology explains the effect. Elliot Aronson's work on self-persuasion shows that conclusions people reach in their own words hold more weight than arguments presented to them. Miller and Rollnick's motivational interviewing, developed in clinical settings, shows the value of normalising a situation and letting the person describe it, rather than labelling it for them. Edgar Schein calls the underlying attitude humble inquiry: asking questions you do not already know the answer to, out of real interest in the other person's view.
Three things tend to follow. People recognise a pattern from their own experience. They attribute it to their own setting, with their own people and workflows. And they elaborate, adding detail no outsider could know: the two engineers who alone can evaluate a new product, one of them retiring next year; the procurement team approving suppliers without seeing quality data from the plant.
That detail is exactly what a deployment needs. It is offered freely, because the conversation is a shared exploration, not an evaluation of a pitch. The buyer stays in control of what they share and what happens next.
Discovery is the first step of the deployment
The information that comes out of these conversations is not sales material. It is the first map of the operation the system will support: the real workflows, the dependencies, the risks and the people whose time matters most. A deployment scoped on that map addresses the actual problem, which is why it is far more likely to reach production.
It also changes the relationship. A buyer who has described their own operation, and seen it understood, trusts the scope that follows. That trust carries into the deployment, when people need to share the exceptions and workarounds the system must handle.
Running a discovery conversation well
Open with patterns, not a product.
Let the organisation choose which pattern fits.
Ask for a recent, specific example, with the people and systems involved.
Explore implications: what does this cost in time, risk or missed revenue, and for whom?
Ask what happens when the usual person is away, which is where hidden dependencies show.
Agree what to bring to a working session: a real piece of work, not a summary.
How Mitochondria's agentic AI systems are scoped
The approach. Every engagement starts with listening. Working sessions begin with patterns seen across sectors and the work the organisation brings, then move into Stimuli, the first deployment stage, which maps the operation as it runs alongside the people who know it.
What it enables. Units configured to the real workflow, its exceptions and its people, with priorities set by the organisation's own description of where time and risk concentrate.
The impact. Deployments that address the organisation's problem, reach production, and earn the trust of the people who use them every day.
Read more about Mitochondria's approach.
Sources
Rackham, N. (1988). SPIN Selling.
Camerer, C., Loewenstein, G. and Weber, M. (1989). The curse of knowledge in economic settings. Journal of Political Economy, 97(5).
Christensen, C. M., Hall, T., Dillon, K. and Duncan, D. S. (2016). Competing Against Luck.
Aronson, E. (1999). The power of self-persuasion. American Psychologist, 54(11).
Miller, W. R. and Rollnick, S. (2013). Motivational Interviewing: Helping People Change (3rd edition).
Schein, E. H. (2013). Humble Inquiry: The Gentle Art of Asking Instead of Telling.
—
Mitochondria builds agentic AI for evidence-led core operations. Research-led and model-agnostic. ISO/IEC 27001:2022 certified, with controls aligned to the GDPR, UK GDPR, India's DPDP Act and Australia's Privacy Act. Amsterdam and Pune. Request a working session →