On Monday morning a trade newsletter reports that the largest rival in your market will quote customers in ten minutes using AI. By nine, three directors have messaged you, each with a different answer: match it this quarter, wait and see, or say nothing and carry on.
Everything in this article is an illustration, with hypothetical numbers. The firm is invented, the rival is unnamed, and none of it is a client story. The method, though, is the one I use when a board asks me what to do about someone else's announcement.
The Monday message
A loud announcement pulls a board into reacting before anyone has read it closely. The message arrives on WhatsApp, the link is forwarded, and by lunchtime the question has quietly changed from "what does this mean for us?" to "why haven't we done this?". Silence starts to look like complacency, and a quick copy starts to look like leadership.
In my experience the costly part is rarely the choice itself. It is that the choice stays private. Each director leaves with a verdict of their own, nothing is written down, and when the next rival announces something the whole debate starts again from nothing. A written verdict, with a reason and a date, is what ends that loop.
An announcement is a claim, not a result
A press line states what a company wants the market to believe. It is not evidence of what your customers will do. So the first job is to read the announcement for what it is, and I use four rungs for that:
- Announced: a statement or a press line.
- Demonstrated: a live or recorded demonstration.
- Available: customers can use it today.
- Reported: customers say they use it, and say what changed.
Treat anything below the third rung as intent. Intent matters, and it deserves a place on the agenda, but it does not deserve the same budget as something your customers can already touch. This is a way of reading, not a statistic, and it says nothing about whether a rival will deliver.
The four questions: the Respond / Watch / Ignore Sort
This is my working method, not research. Put the four questions on one page and answer them before the board meets.
- What exactly is announced, and on which rung? Name the rung in writing. If the honest answer is "announced" or "demonstrated", say so.
- Which of our customers could this change? Name them. Count the accounts and their share of revenue from your own last twelve months of data, not from the rival's slide. If you cannot name the accounts, the answer so far is noise.
- What does it change: customer behaviour, their cost, or our channel? A faster quote changes how buyers behave, and that can bite quickly. A cheaper back office changes the rival's price later. A new channel changes where buyers look. Only the first usually forces your hand within months.
- What does being late cost, and what is the smallest step that caps it? Estimate the weeks of delay, the gross profit of one exposed account, and the cheapest response that protects the exposed accounts only.
Then the verdict, always written down, always one of three:
- Respond: a small, dated, funded step aimed at the named accounts.
- Watch: two observable tripwires and a review date.
- Ignore: the reason, in one sentence, so the next director's message meets a written answer.
A worked example: a Beirut distributor
Here is the sort applied to an invented firm. Every figure below is a hypothetical number for illustration. The firm is a mid-size Beirut distributor of industrial supplies with annual revenue of $20,000,000 and a gross margin of 20%. Its largest rival announces AI-assisted quoting in ten minutes. Ours takes about two days.
Question 1. The rival has a press line and one demonstration video, and no customer has reported using it. That is the second rung, demonstrated, and not yet available.
Question 2. Of our 40 largest accounts, 6 have asked for a same-day quote at least once in the last twelve months. Those six are 6 / 40 = 15% of accounts. Assume they account for 24% of revenue:
- 0.24 x $20,000,000 = $4,800,000 a year across the six
- $4,800,000 / 6 = $800,000 per account, on average
- 20% x $800,000 = $160,000 gross profit per average exposed account, per year
Question 3. A faster quote changes buyer behaviour, and only for those six. Assume the other 34 accounts buy on relationship and delivery reliability.
Question 4. Assume the smallest useful step is a fast-quote pilot for the six accounts only: 8 weeks, $60,000. Set against one account's gross profit, $60,000 / $160,000 = 0.375 of a year, and 0.375 x 12 = 4.5 months. Put plainly, the pilot is paid for if it keeps one exposed account for about four and a half months longer than it would otherwise have stayed. That is a comparison, not a guarantee.
| Verdict | What it covers | What is written down |
|---|---|---|
| Respond | The six exposed accounts | Fast-quote pilot, $60,000, 8 weeks |
| Watch | The other 34 accounts, and the "AI-powered everything" promise | Tripwire 1: two of the six ask us to match a quoted turnaround. Tripwire 2: the rival's offer appears in a tender we are in. Review in 30 days. |
| Ignore | The rival's "AI-first company" brand claims | No customer or cost effect is shown, so no budget |
Notice what the sort did not do. It did not match the rival across the whole business, and it did not shrug. It aimed a small amount of money at 15% of accounts, and gave the remaining questions a date.
A caution on the inputs. The 24% revenue share, the 20% margin, the $60,000 and the 8 weeks are assumptions, not research. The sort does not predict what a rival will do. What it does is make the decision visible, and cheaper to revisit.
What goes in the board minute
The minute can be four or five lines. For the example above it would read:
Rival AI quoting: on the demonstrated rung, not yet available to customers.
Respond: fast-quote pilot for six named accounts, $60,000, 8 weeks.
Watch: remaining accounts; tripwires are two of the six asking us to match a quoted turnaround, or the rival's offer appearing in a tender we are in.
Ignore: the "AI-first" brand claims, no customer or cost effect shown.
Review date: 30 days.
When the next announcement lands, the director who messages you meets a page, not an argument. If a tripwire trips, the board knows in advance what it said it would do. If it does not, the review date forces an honest look rather than a drift.
A practical note on preparation: ask one person to gather the account numbers from your own sales records, with a deadline of three working days. The sort is only as good as the twelve months of data under Question 2.
Monday action: pick the last AI announcement from a rival or peer that unsettled your leadership, and write its Respond, Watch or Ignore verdict on one page this week. If you cannot name the accounts it touches, the answer so far is noise. If you would like a second pair of eyes on your board's AI roadmap, write to me.
Related evidence: Wikipedia defines vaporware as a product that is announced to the public but is then significantly delayed, never released, falls short of expectations or is cancelled, which is why an announcement on its own is not evidence that a capability exists. (the long record of announced products that shipped late or never)
Suarez and Lanzolla call first-mover advantage a half-truth: executives almost universally believe early entry gives an insuperable head start, yet the fate of the dot-coms, its most convinced adherents, is a cautionary lesson. (Suarez and Lanzolla's Harvard Business Review analysis of first-mover advantage)
