Good projects fail all the time because someone important wasn't asked. Stakeholder analysis is how you find that person before they find you.
What it is
A stakeholder is anyone who affects, or is affected by, a decision. R. Edward Freeman made the idea central to management thinking in 1984. A common tool is the power–interest grid, often credited to Aubrey Mendelow:
| Low interest | High interest | |
|---|---|---|
| High power | Keep satisfied | Manage closely |
| Low power | Monitor | Keep informed |
For each stakeholder, you also note: what they want, what they fear, and what they'll likely do.
Why it matters for bots
Bringing in bots changes people's jobs. That's a stakeholder problem more than a technical one.
- The estimator whose knowledge the quote bot depends on.
- The office manager who now approves bot drafts.
- IT, who has to approve the connectors.
- Customers, who may have opinions about bot-written emails.
- The person who used to do the job the bot now does.
Skip them and the bot works fine technically, and fails anyway.
Why people skipped it
It feels political, and it takes time to think through everyone. Teams go straight to the plan.
How a bot does it
- Give the bot the plan, the process map and the org chart.
- Have it list everyone affected, by each step of the process.
- Have it place them on the grid, with a reason.
- Have it draft what each wants and fears. These are guesses, and marked as such.
- Have it propose how to involve each one.
- A person who knows the people corrects it. This is where real knowledge matters most.
The bot brief
Draft a stakeholder analysis for [change].
Inputs: plan [file], process map [file], org chart [file].
List everyone affected, step by step through the process. Include outside parties
(customers, suppliers, IT, auditors).
For each: role, power (H/L) and interest (H/L) with a one-line reason, what they
likely want, what they likely fear (mark both GUESS), and a suggested way to involve them.
Place them on a power-interest grid.
Flag anyone whose job content changes because of this. Don't speculate about
individuals' personalities; stay with roles and effects.Starting point only. Test it on your own data and model, and adjust as needed before you rely on it.
An example
Change: quote drafting moves to a bot.
| Stakeholder | Power / Interest | Wants | Fears (guess) | Involve by |
|---|---|---|---|---|
| Senior estimator | High / High | Fewer late nights | Being replaced; errors under their name | Write the pricing rules with them; they approve quotes |
| Sales manager | High / High | Faster quotes | Lost deals from wrong prices | Show test results before go-live |
| IT | High / Low | No security problems | Unknown connectors | Review tool access up front |
| Customers | Low / Medium | Fast, accurate quotes | Impersonal service | Person signs every quote |
The first row is the one that matters most. The estimator's knowledge is the skill. Their support decides whether this works.
Where bots differ
A bot can list roles thoroughly. It can't know that the plant manager and the sales manager don't get along, or that the estimator is retiring next spring. People fill that in. Keep the bot's draft about roles, not personalities.
What this means for you
- Before bringing in bots, list everyone whose work changes.
- Have a bot draft the grid.
- Correct it with what you know about the people.
- Involve the people whose knowledge the bot depends on. Early.