Most AI conversations begin with what a tool can do.
Can it read the tender pack? Can it pull together a first draft? Can it help a stretched team deal with the pile of documents that has arrived just before a deadline?
Those are sensible questions. If you are leading a growing developer and the same few people are covering auctions, procurement and financial close, they are often the only questions anyone has time to ask.
But I think there is a more useful place to start: which commercial date or decision would this actually improve?
An AI pilot can be technically impressive and still leave the business in exactly the same position. The team may produce a first draft faster, but still struggle to decide which opportunity to pursue, compare supplier returns, carry commitments into contract negotiation or answer the next lender question with confidence.
That is why I am wary of starting with a technology shopping list. The better starting point is the live pressure in the work.
Start with the date that matters
For one team, it may be an auction or tender window. The work has to become coherent quickly enough for people to focus on the offer, the risks and the decisions that will actually change the result.
For another, it may be procurement moving into financial close. Supplier returns need to be compared, qualifications understood and commitments kept visible as commercial, technical, legal and finance colleagues each take their turn with the same project.
Sometimes the pressure is less visible. A board asks for an AI plan, while people are already using ChatGPT or Copilot individually and nobody can say which sources they used, what was checked or where a person is meant to make the call.
These are different situations, but they have something in common. The problem is not usually a lack of possible AI tasks. It is that the work has become hard to move through a small number of people, documents, decisions and deadlines.
Two routes worth separating
The first route is to improve commercial work that already exists.
AI can help a team find requirements, compare documents, prepare a first view of risk or gather the evidence behind a local-content or bankability claim. Used well, that does not make the commercial judgement automatic. It gives the people making the judgement a better starting point, with sources, open points and exceptions visible.
The aim is not simply to make a bid look finished sooner. It is to create more room for the work that still needs people: understanding the customer, challenging an assumption, negotiating a position, deciding which risk to accept and owning the commitment.
The second route is different. It asks whether a business has expertise, a repeatable internal process or a body of insight that customers might value in another form.
That can be exciting, and it can also become a distraction very quickly. Not every internal tool deserves to become a product, and not every useful piece of analysis has a viable customer market behind it.
Before getting carried away, I would test five things:
- Is there a real customer pain?
- Do we have a genuine right to win?
- Can it be delivered safely and consistently?
- Is there enough margin once the human work is counted honestly?
- Can we genuinely sell and support it?
Those questions are deliberately commercial. They stop a team from treating an interesting capability as proof of an offer.
What makes either route useful
Both routes need more than a good prompt. They need a clear job, reliable context and named human responsibility.
The system may prepare work, compare evidence or coordinate a bounded part of the workflow. It should not become the place where assumptions, commitments and risk decisions disappear because nobody can see what happened between the source material and the answer.
That is where governance becomes practical rather than bureaucratic. A commercial leader does not need a large control framework before testing anything. They need to know which information is trusted, what the system is allowed to do, what still needs human challenge and who is responsible for the final decision.
Without that, a pilot may produce a polished output but leave the team with more checking, more uncertainty and another tool that does not quite fit the way the work really moves.
A more grounded first step
Do not start by trying to automate the whole commercial function.
Start with one live pressure: an upcoming auction, a difficult procurement package, a project approaching financial close, a weak evidence trail, or a new offer worth testing.
Map the people involved, the information they need, the decisions that have to be made and the constraints around them. Only then decide where an AI-enabled workflow could genuinely help.
That keeps the exercise anchored in the work, rather than the technology.
The question is not whether AI can produce something.
It is whether it can help your team move the decision that matters, while leaving the right people able to see, challenge and own the result.
About the author
Yorik Tisseau (MBA) is the founder of Ignis Leadership. He has 17 years of engineering, operational and commercial experience across energy and infrastructure, including bp, RWE and Bouygues, from project engineer to Bid Director.
Today, he helps growth-stage developers and IPPs design governed human-AI workflows around the commercial decisions and deadlines that matter.
