Book a call
Menu
Field Guide · level 6 · Advanced and by role

Why don't better AI tools alone change results, and what should a manager do?

20 min · as of Oct 2026
You will be able to

Explain why the organisation moves results more than the tools do, and take five manager actions: use AI in the open, redesign a few workflows, say what to do with time saved, set the rules, and measure outcomes.

Part of Lead the changeTime about 20 minutes, with the try-itChecked 16 Oct 2026

Three things to take away

  1. 1

    In the 2026 surveys, how the organisation leads and organises the work is linked to far bigger reported gains than individual habits, and gains for individuals have not yet moved company results.

  2. 2

    A manager's job is to change the work around the tool: show your own use, redesign two or three workflows, and say what the time saved is for.

  3. 3

    Measure outcomes against a starting point, not hours saved. And read every survey, including these, for who ran it and what it can show.

In every assistant · as of Oct 2026

In your tool

The five actions are the same in every assistant, so they work in whichever one your company already uses. A few features in each make them easier. As of Oct 2026, checked on the vendors' own pages on 16 Oct 2026; what you see depends on your plan and your admin.

For action 2, redesign a workflow:

  • Give the workflow one home: a shared project with its files and its instructions. In Claude, Team and Enterprise plans can share a project with "Can view" or "Can edit". In ChatGPT, every plan can share a project with chat or edit access; a shared project always uses project-only memory, and Work is not available in it. Gemini has no projects by that name; the nearest thing is a notebook.
  • Write the redesigned steps as a skill. A skill is a saved how-to: one SKILL.md file with a name, a one-line description and the steps. Claude, ChatGPT and Gemini all accept the same format, so one file can serve a team that uses more than one tool. Test it in each. In Gemini, skills are on personal accounts now, and on work and school accounts "soon".
  • Schedule the part that repeats. Claude has scheduled tasks, such as a report that starts every Monday. ChatGPT Work has Scheduled Tasks that can also start when a new Gmail or Slack message arrives. Gemini tasks can run on a schedule, on personal accounts with Google AI Pro or Ultra.

For action 4, set the rules: see "What rules does your team need for AI?", which shows where an admin sets them in each assistant.

For action 5, measure outcomes:

  • Claude's Dashboards (beta, Pro, Max, Team and Enterprise) build live charts from connected data, and each one shows its query and when it last refreshed.
  • ChatGPT's Data plugin analyses connected business data and builds dashboards in Work.
  • Gemini Notebook can now run code to analyse data in your sources (Ultra and Workspace business editions first).
  • Whatever builds the chart, check the starting point and the measure yourself.

Choosing the tool for your team

The team's main needPick
Documents, research, spreadsheets and handed-over tasksClaude
Routines that start when an email or Slack message arrivesChatGPT Work
A company that already runs on Google Workspace (Gmail, Docs, Meet)Gemini inside those apps
A company that has standardised on one toolThat tool

My recommendation: Claude first, ChatGPT second, Gemini third.

For a team, Claude is the simplest to roll out: one conversation for questions and handed-over tasks with no mode to choose, it asks before acting by default, finished documents on every plan, and one Customize page for connectors, skills and plugins. ChatGPT is a strong second, with Work and its event-triggered tasks. Gemini is the natural fit for a Google Workspace company, but its agent tasks, custom apps and skills are limited to personal accounts for now.

And this lesson's whole point applies to my ranking too: the organisation moves results more than the tool does. If your company has standardised on one tool, use it, and spend your energy on the five actions.

On your own work · nothing private

Try it (10 minutes)

Use one job your team really repeats, such as a weekly report, a customer reply or a monthly pack. You will plan on paper; paste nothing into any AI tool.

  1. The job today (2 minutes). Write its steps in a line each, and who does each one.
  2. The redesign (3 minutes). Mark the one step AI could draft, and the step where a named person checks it.
  3. The starting point (2 minutes). Pick two measures: one for quality (errors, rework, corrections) and one for speed (hours, days to send). Write today's number, or how you will find it.
  4. The time saved (1 minute). Write one line: what the team will do with the time.
  5. In the open (2 minutes). Write what you will show at your next team meeting: one piece of your own work, how AI helped, and what you checked.
Three questions · pick one answer each

Check yourself

  1. 1In the 2026 Work Trend Index, organisational factors accounted for more than twice the reported AI impact of individual factors. What does that tell you?

    The answer: C Microsoft tested factors against self-reported AI impact. It is a strong association in survey answers, and individual factors still counted for 32 percent.

  2. 2In McKinsey's 2026 survey, how did the "high performers" differ from everyone else?

    The answer: C They redesigned the work rather than adding AI to the old steps, and were twice as likely to say their senior leaders show commitment and that they measure impact. They are about 6 percent of respondents, and the answers are self-reported.

  3. 3Your team says AI saves them time every week. What is the manager's next move from this lesson?

    The answer: C Time saved is not value until it shows up in an outcome. In McKinsey's survey, 80 percent say AI improved their individual productivity, yet only 37 percent attribute any EBIT impact to it.
0 of 3 answered
Read the whole lessonThe idea in full, the worked examples, common mistakes and what to check

The idea

AI is paying off for individuals. In McKinsey's 2026 survey, 80 percent of respondents say AI has improved their individual productivity. But, in McKinsey's words, "These individual gains have yet to translate into broad financial impact for organizations." Three recent studies point the same way: the gains follow how the work is organised and led, more than the tool.

What the organisation does. In Microsoft's 2026 Work Trend Index (a survey of 20,000 knowledge workers who use AI at work, in 10 countries, run by Edelman Data x Intelligence between February and April 2026), factors such as culture, manager support and talent practices "account for more than 2x the reported AI impact of individual factors like mindset and behavior (67% vs. 32%)". The same report found that only 26 percent of AI users say their leadership is "clearly and consistently aligned on AI".

What the manager does. A separate Microsoft-led study of 1,800 workers (July 2025, people from outside Microsoft, most of them leaders and managers) found that when managers actively modelled AI use, employees reported a 17-point lift in AI value (a score made from their own survey answers).

Redesign, not sprinkling. In McKinsey's State of AI 2026 (published 25 Aug 2026; an online survey of 1,719 participants in 97 nations, May to June 2026), 37 percent of respondents attribute at least some EBIT (earnings before interest and tax) impact to AI use, about the same share as a year before. The "high performers", about 6 percent of respondents, who attribute at least 5 percent of EBIT to AI and call its impact significant, stand out: "Nearly three-quarters of high performers report fundamentally redesigning workflows because of their AI use", against "just one-quarter of other respondents". They were also twice as likely to say their senior leaders show commitment to AI and that they have defined ways to measure its impact.

Transform, not just adopt. McKinsey's own reading of its survey: the organisations that turn individual gains into lasting financial results "are likely to be those that transform their businesses, not just adopt AI tools". A commentary beside the survey, by one of its authors, puts it more plainly: "The limiting factor is increasingly the organization's ability to absorb change." Both are McKinsey's view, not a measured finding.

Read these with care. These three studies come from two companies that sell AI tools or advise on AI: Microsoft sells AI tools and McKinsey sells consulting. Every figure is what people said in a survey, not a measurement of their work. And they show links, not causes: Microsoft's own report says it tested factors "against self-reported AI impact". Use them as a direction, not as proof, and measure your own team.

The five manager actions

1. Use AI in the open. Show your team one real piece of your own work where AI helped, and what you checked before you used it. In Microsoft's study, managers modelling AI use went with a 17-point lift in the value employees reported.

2. Redesign two or three workflows. Pick a few jobs the team repeats and redesign each from start to finish: what AI drafts, what a person checks, what changes hands, and what stops being done at all. Do not spread AI thinly over every task.

3. Say what the time saved is for. If AI saves someone an hour, decide together where that hour goes: more customer time, faster turnaround, the backlog, learning. If nobody decides, nobody can say later where the hour went. McKinsey's survey shows the same gap across organisations: most respondents report better individual productivity, while the share reporting any profit impact barely moved.

4. Set the rules. One page: the tools, what never goes in, who checks, what agents may do without asking, who to tell. That is the lesson "What rules does your team need for AI?"

5. Measure outcomes, not hours. Before you change a workflow, write down where it stands today: quality, rework, turnaround, errors caught and how confident people feel. Check the same things after a month. Time saved is not value until it shows up in one of those.

An illustration, not a real case: a manager picks the weekly client update. Today it takes a team member most of a Friday morning and often goes out with an old figure. The redesign: the assistant drafts it from the shared tracker on Thursday, the account lead checks every figure against the tracker, the manager reads it once. They agree the freed morning goes to client calls, and they track two things: hours to send it, and corrections clients ask for.

Common mistakes

  • Buying licences and waiting. In these surveys, individual gains alone have not moved company results.
  • Switching tools to fix results. In these surveys, how the work is led and organised counts for more than the tool. Pick one approved tool and change the work around it.
  • Using AI quietly while asking the team to use it openly. Show yours.
  • AI on every task, redesign of none. Pick two or three workflows and change them properly.
  • Counting hours saved as the result. Hours are an input. Quality, speed and rework are results.
  • No starting point. Without today's numbers you cannot tell whether anything improved.
  • Rules after the first incident. Agree them before the first agent runs.
  • Quoting vendor surveys as proof. They are a direction, from companies that sell AI, based on what people said.

What to check before you trust it

Any claim about what AI did for results, including the ones in this lesson:

  • Who ran the survey or study, and do they sell AI tools or advice?
  • How many people, where, and when? Is it what people said, or what was measured?
  • Does it show a cause, or only a link?
  • Is "impact" defined, and is it the same thing you care about?
  • For your own team: did you measure before and after, on the same measures?
  • If an AI tool summarised the study for you, open the study itself. Quotes and links can be invented too.

Check it. Before you report a gain, show the starting point and the measure.

Keep it safe. Set the rules before the first agent runs, not after the first slip.

Checked against · read 16 Oct 2026Every claim, its source and the date read

As of: product steps as they were on the vendors' own pages on 16 Oct 2026.

Every figure above was read at its source on 16 Oct 2026.

The illustration in this lesson is made up. Neosaipien is not affiliated with Anthropic, OpenAI, Google or Microsoft. Product names belong to their owners.

Where next

Next lesson · level 6 · 20 min

What rules does your team need for AI?

Agree a one-page set of team AI rules, and know which 2026 laws it touches (not legal advice).

Open the next lesson
The kit · free download

The AI Rollout Plan (30-60-90 days), for leaders deciding how a whole organisation takes up AI: the plan, a plan on a page, a job picker and measures tracker, and a leadership deck.

Bring this to your team

Executive coaching. One-to-one sessions on a leader's own work, until AI is part of how they do it. Also the Leadership briefing, for your leadership team, and AI implementation.

Book a 20-minute call

Back to level 6, Advanced and by role