For owners, CEOs and boards
Questions a board will ask about your AI plan.
The board packet goes out on Friday. Item six reads: "What is our AI plan?" The draft answer is a list of tools. This page lists the eight questions a board is likely to ask next, and what a good answer to each one contains.
ILLUSTRATIVE SCENARIO. The packet above is an example, not a described client.
How to use this list
Ask each question out loud, a week before the meeting.
Put each question to the person who would have to answer it. A good answer is specific to your company, has a named owner, and says what happens when something goes wrong. These are the weak answers to listen for: "we are looking into it", "IT handles that", and "we bought everyone a license".
The eight questions
What a board will ask, and what a good answer contains.
Each answer should be in plain language and owned by a named person.
Where does AI change the economics of our industry: price, cost, speed, or who wins the customer?
A good answer names one or two places in your own industry where AI moves price, cost, speed or customer choice, and says which of them reaches your margins first. It also says where AI changes nothing for you, so the board is not asked to react to everything.
What would our company look like with AI doing the routine preparation, and how far are we from it?
A good answer describes ordinary work, such as quotes, invoice matching and the weekly numbers, with AI doing the gathering and drafting and a named person making each decision. It then says in plain terms how far today's company is from that picture and what stands in between.
Which three moves matter most in the next twelve months, and which can wait?
A good answer names three moves, each with an owner and a measure, and says what is being left for later and why. A list of ten equal priorities is a sign that nobody has chosen.
What should we build, buy, partner for, or skip?
A good answer sorts each item into build, buy, partner or leave alone, with a reason, and starts from the work to be done rather than from a product. Skipping something for a stated reason counts as a decision.
Who owns the plan, and how will the board see progress?
A good answer names one person who owns the plan and what happens the day after it is approved, not a committee. It also says what the board will see each quarter and which number moves if the plan is working.
What may a system do on its own, what needs a person's approval, and how do we stop it?
A good answer writes down three lists: what a system may do alone, what needs a named person's approval, and what it must never do, such as sending money or making a promise to a customer. It also says who can switch a system off and how quickly.
What information leaves the company, to whom, and under what terms?
A good answer says which records are going into outside tools, who has access, and what the written terms allow the provider to do with them. Ask the people who do the work, not only IT, because some of this use may never have been decided.
What will it cost, and how will we know it worked?
A good answer gives a cost range and what it includes, plus a baseline taken before anything starts: margin, response time, cash, errors or capacity. If a change cannot be measured against that baseline, it should not be in the first build.
Who wrote this
Darren Weiss, founder of 013 Labs.
An operator who helped scale a regional company into a major participant in a highly regulated national industry, a lawyer by training, who designs, builds and runs AI systems every day. This page gives general guidance, not legal, tax, accounting or investment advice.
Start with the one-page guide.
The first answer most boards want is who in the company already uses AI well, and what they must never build. The guide shows how to find that person and what to write down first.