The Four Discovery Pulses
The four Discovery pulses
Four one-off pulses for an organisation that hasn't started with AI, or has only just begun. They're described together because you run them together: two that tell you where you are, and two that tell you where the gains might be.
Everything they have in common is covered in taking your AI baseline: how a one-off opens and closes, what people can and can't skip, how they're reminded, and how to read the Baseline Report. This article is what each one asks and what you get back.
What they share
- Everyone gets them. All four are org-wide, so they reach every member of your organisation including the people no AI tool has been assigned to. On a pre-adoption baseline that's most of them, which is rather the point.
- Once each. One-offs, about three minutes, five or six questions.
- Required questions can't be skipped, and the open questions are all optional. Nobody is made to write prose.
- They can't disturb your existing numbers. None of them is attached to an AI tool, so nothing they collect touches adoption, satisfaction, hours saved or ROI for any tool you're already measuring. A baseline taken before adoption can't move the numbers adoption is measured by.
- Each carries one gauge, a single 0-100 reading you can quote now and re-run in six months.
- Pro, like everything else in the pulse library. Free trials run on Pro.
The baseline pair
The Current AI Use Pulse
The job: what people already use, whether it helps, and where they got it. Run this one first.
What it asks:
| # | Question | What it gives you |
|---|---|---|
| 1 | Which kinds of AI tool have you used for work in the past week? (pick any) | The real footprint: chatbots, coding assistants, AI inside tools you already own, image and design, notetakers, or none |
| 2 | In a normal week, how often does AI help you get work done? | Frequency, from daily to not at all |
| 3 | Do you use AI tools your organisation hasn't provided? | The shadow-AI figure |
| 4 | Overall, how useful is AI to your work right now? (1-10) | The Usefulness gauge |
| 5 | How did you learn to use the AI tools you use? | Whether your AI learning is doing anything, or whether people are teaching themselves |
| 6 | Which AI tools do you actually rate, and what for? (optional) | Themed list of what people genuinely value |
What to look for. Question 1 against your licence list is the fastest useful thing in the whole set: tools people use that you don't pay for, and tools you pay for that nobody named. Question 5 usually settles the "do we need AI learning" argument on the spot, because "worked it out myself" leading is not a sign of a healthy rollout, it's a sign nobody was shown anything.
The AI Risk Pulse
The job: what people understand about the rules, and what has already gone into public tools. It's a temperature check, not an audit.
What it asks:
| # | Question | What it gives you |
|---|---|---|
| 1 | Do you know what your organisation allows? | Policy awareness, from "yes, it's clear" to "I didn't know there were rules" |
| 2 | Have you put work information into a public AI tool? | How widespread it is |
| 3 | What kind of work information have you put into a public AI tool? | Nothing sensitive, internal, or client and personal |
| 4 | How often do you check AI output before you use it? | The Verification gauge |
| 5 | What would make it easier to use AI safely here? (optional) | Themed list of what people say they need |
Two things about how it's built. Question 3 is asked of everyone rather than only of the people who answered yes to question 2, because "nothing sensitive" is a true answer for someone who hasn't done it, and asking everyone keeps the denominator the whole organisation rather than only the people who owned up.
And the answers are anonymous, which is the only reason the numbers are worth anything. People who think a risk question is a compliance exercise answer it like one. Everyone receiving it is told their answers are anonymous before the first question, and the report never shows who said what. If you plan to talk to your team about this pulse beforehand, say the same thing.
What to look for. The pairing of questions 1 and 4. High "roughly" or "no" on the rules with low Verification is not a discipline problem, it's a missing policy nobody has published. Question 5 usually tells you which one.
The opportunity pair
The AI Opportunity Pulse
The job: where the week actually goes, and what people would hand over first. This one's value is in the writing, not the charts.
What it asks:
| # | Question | What it gives you |
|---|---|---|
| 1 | What's the most repetitive task that eats your week? (optional) | Themed list, in people's own words |
| 2 | Roughly how much of your week goes on routine, repeatable work? | The size of the prize |
| 3 | Which of these take up the most time in your week? (pick any) | Finding information, writing, copying between systems, chasing people, reporting |
| 4 | If something could take one task off your plate entirely, which would you choose? (optional) | Themed list: your candidate use cases, ranked by how many people asked for them |
| 5 | How keen are you to use AI for more of your work? (1-10) | The Appetite gauge |
What to look for. Question 4's theme list is the deliverable. It is a use-case backlog written by the people who'd use it, ranked by demand, and it beats a workshop because nobody in it is performing for their manager. Read it beside question 3: if "copying between systems" leads the chart and the open answers are all about the same two systems, you have found your first project without doing any analysis.
The Agent & Automation Pulse
The job: which repetitive work has the shape software could take on, and what's genuinely in the way.
What it asks:
| # | Question | What it gives you |
|---|---|---|
| 1 | Which of these describe work you do regularly? (pick any) | Same steps in the same order, moving data between systems, checking or approving, answering the same questions, drafting from a template |
| 2 | Describe one task you could hand to a capable assistant, step by step. (optional) | Themed list, with enough detail to scope from |
| 3 | For the most repetitive work you do, where does the information you need live? | One system, a few, email and chat, people's heads, or paper |
| 4 | What would stop you handing your most repetitive work over today? | The blocker |
| 5 | How comfortable would you be handing your most repetitive work to software? (1-10) | The Readiness gauge |
What to look for. Question 3 is the feasibility question dressed as a survey question. Work whose information lives in one system is buildable now; work whose information lives in people's heads is a documentation project first, whatever anyone promises you. Question 4 tells you which conversations are technical and which are about trust, and Readiness tells you how hard the second kind will be.
Why questions 3, 4 and 5 don't refer back to question 2. They used to say "that work", which meant three required questions leaned on an optional one. Someone who left question 2 blank was then made to answer three questions about something they'd never written down. Each one now names its own subject.
Which to run, and when
| If you want to | Run |
|---|---|
| Know where you stand before spending anything | Current AI Use, then Risk |
| Build the case for a programme | The baseline pair, then the opportunity pair a fortnight later |
| Answer "is anyone actually doing anything risky" | Risk on its own |
| Find the first automation project | Opportunity and Agent & Automation together |
| Show movement six months in | The same pulse again, and put the two reports side by side |
Running all four in one week asks about twelve minutes of everyone's time and usually costs you completion on the last two. Two at a time, a fortnight apart, holds up better.
Related: Taking your AI baseline · The pulse library · Who sees my answers?