From ChatGPT Experiments to a Shared Team Workflow

Someone on your team has found a useful way to work with AI. It helps them prepare for a meeting, work through research, or put together a first draft. That’s a useful starting point. What would it take for someone else on the team to use it well?

Start with what makes their approach work. What information do they give it? What do they check? Where does their experience help them spot something that needs changing? Those details matter when another person takes on the work.

You may already have much of what you need. The next step is to make the approach clear enough for others to use, agree where a person’s judgement is needed, and decide who will keep it useful as the work changes.

Useful individual practice does not automatically become a shared team capability. The difference shows up in delivery: one person produces steady work, another gets uneven results, and the business cannot yet point to how that kind of work is supposed to get done.

When personal practice stays personal

A useful prompt often lives in one person’s chat history. When they are busy or away, colleagues start over or copy something half-remembered. Review is informal. When the work goes out, it can be hard to say what “good” meant, who checked it, or who will update the approach as the work changes.

That pattern is common while people are still learning what helps. It is not a failure of initiative. It becomes harder to live with when clients, colleagues, or partners depend on the output, and the business still cannot explain the shared way of working.

You will also see a second version of the same problem: two people both “use ChatGPT well,” but they prepare different inputs, check different things, and mean different things by done. The tool is shared. The practice is not.

Personal practice and team practice

Use this as a check, not a scorecard of failure. If most answers sit on the left for a given piece of work, the practice is still individual. If you can answer the right-hand column for one recurring workflow, you are building something the team can share.

You do not need every column perfect on day one. You need one workflow where the team can say what goes in, what comes out, where human judgement matters, and who keeps the instructions useful.

What to make clear before you share it

Before you ask everyone to use the same tools the same way, make the work itself clearer:

1. What people need to know – The facts, examples, constraints, and context the model and the person need for this kind of task. OpenAI’s guidance on clear prompts and context is a useful starting point.
2. Who does what – Who prepares inputs, who runs the assist, who reviews, who delivers.
3. Where AI can help – Be specific: drafting, summarizing, structuring, checking against a list.
4. Where human judgement matters – Choices that affect clients, quality, risk, or reputation stay with a person who is responsible for the result.
5. Quality checks – What “done” means, and what gets reviewed before the work leaves. OpenAI’s guidance on checking ChatGPT’s answers explains why important information needs verification.
6. How the work moves – How a request becomes a finished piece without living only in one person’s chat.
7. Who keeps it useful – Who updates prompts, instructions, and examples when the work changes.

In plain terms, this is how people and AI work together on real delivery. It is not a tip sheet for better prompts.

A simple flow is enough to start:

Request → prepare agreed inputs → AI assist → human review where judgement matters → deliver → improve the instructions from what broke.

If you cannot draw that flow for one recurring task, the team is not ready to standardize the tools around it.

A simple first shared workflow (hypothetical)

The following is a labeled hypothetical, not a client result.

A ten-person specialist firm has several people already using ChatGPT to draft client updates. Partners value the speed when it works. They also notice uneven tone, missing context, and last-minute fixes.

They choose one recurring task: the weekly client status update for active projects. They agree on:

• Inputs: project notes from the shared tracker, open risks, decisions made this week, and anything the client asked for.
• AI assist: a draft built from those inputs using a shared instruction note, not a private chat.
• Human review: the project lead checks facts, tone, and anything that could commit the firm, then sends.
• Done: the update is accurate, readable, and does not invent status.
• Owner: one delivery lead keeps the instruction note current and notes what went wrong each week.

The work runs like this: a request comes in; someone prepares the agreed inputs; AI helps with a draft; a person reviews where judgement matters; the update goes out; the instruction note improves from what broke.

After a few cycles, they are not simply “using ChatGPT more.” They have a team way of running one kind of work. That is the shift that matters.

Before you expand beyond that first workflow, check three things. Are the inputs written where the team can find them? Is there a named review step where human judgement matters? Is there an owner for the workflow itself? If any answer is no, settle those before you add tools or train the whole company.

Read the diagram as text

Gather project notes from the shared tracker, including risks, decisions and client requests. A team member prepares the inputs. AI drafts the update using a shared instruction note. The project lead checks facts, tone and commitments, then sends the reviewed update. Record what went wrong; the workflow owner updates instructions for the next cycle.

Where teams often get stuck

Too many tools, too little standard. People pick different apps and prompt styles, and save work in different places. The business still cannot see one clear approach.

Speed without review. Output moves faster, but there is no check where mistakes matter. It becomes unclear who is responsible for the result.

A private method and a public one. The approach that actually works stays in one person’s habits, while the documented version gathers dust.

Training without redesigning the work. A workshop helps people try things. On its own, it does not create shared inputs, review, or ownership.

Building before the work is clear. Automating an unclear process can lock the confusion in. Get clear on how the work should get done, then decide what to build.

Why the handoff matters

A clear prompt is only one part of a shared workflow. The handoff needs just as much attention.

Another person also needs to know which sources count, what “good” means for that client, and who is allowed to send the work. Without those details, the shared way of working is incomplete.

Documentation needs upkeep as the work changes. Name an owner for the workflow – not only for the task – so someone is responsible for keeping the shared instructions useful.

That is why the checklist above starts with inputs, review, and ownership, not with a longer prompt library.

Design, Implementation, and Assessment

If you already know which practice to share, and you need help defining roles, where judgement matters, and how the work should run, that is Design work – sometimes with Implementation if you also need the systems connected and tested on real tasks.

If enough clarity already exists to build and put the approach into practice, you can move toward Implementation directly.

If the team cannot yet choose which workflow to share first, an Assessment can help decide where AI could make a worthwhile difference and what to do first. Use it when you need that decision. It is not a compulsory first step.

Mode Lab helps teams improve how their business works – and build new capabilities – with AI. We bring your team’s expertise and AI together to rethink how the work gets done, and help you put it into practice. You can engage us for Design, Implementation, or both.

If you want to build capabilities yourself through learning and practice, Mighty AI Lab is a separate learning and community pathway for individuals and teams.

Start with one useful practice

Pick one useful individual ChatGPT practice your team already relies on – whether it improves work you already do or supports something new you want to offer. Write down the inputs, where human judgement matters, and who will keep the approach useful. Run it the same way for a few real cycles. Improve the instructions from what breaks.

When you want help turning that into a shared way of working, or deciding where to start, start a conversation. Tell us one workflow people already run that should become a team standard.