You want to grow the business, but the team is already balancing client work, sales and delivery. ChatGPT can produce a convincing growth plan from that description. The harder question is whether the plan fits your customers, your capacity and the money you can afford to put at risk.
Use ChatGPT to prepare and challenge one business decision at a time. Give it the relevant evidence, ask it to compare realistic alternatives, then examine what would have to be true for its recommendation to work. The result should be a short decision record: what you have chosen, why, what remains uncertain, who owns the next step and when you will review it.
You can follow this process with an approved ChatGPT workspace and a small set of business records you are authorized to share. Pasted summaries are enough to begin. You do not need to build a chatbot or train an AI model. You do need someone who understands the business and has authority to make the eventual choice.
What ChatGPT can contribute to strategic planning
ChatGPT is an AI assistant you can ask to summarize information, suggest alternatives and organize a comparison. In planning, that gives you a way to work through customer feedback, question an assumption or prepare a discussion with colleagues. Giving it context helps specify the task; it is not the same as training or fine-tuning the underlying model.
The important boundary is evidence. An answer may sound confident while containing incorrect facts or fabricated references, as OpenAI explains in its guidance on ChatGPT’s limitations. Treat a generated recommendation as something to examine. Ask where the relevant claims came from and open those sources. Check the date, read the relevant passage and confirm that it supports the claim in your context. If a reference is missing or cannot be checked, leave the claim unresolved.
The example below follows a fictional design studio choosing its next growth move. All company details, records and proposed tests are illustrative. They show how to use the method, not the results of a client engagement or a tested prompt sequence.
1. Define the decision and its constraints
Start by writing the choice you need to make. “Help us grow” leaves the goal, time frame and acceptable trade-offs open. “Should we offer monthly support to existing clients or develop a fixed-price workshop for new clients?” gives the discussion a shape.
For the fictional studio, the objective is to develop a source of recurring work without hiring this quarter or delaying existing projects. The owner is deciding which offer to test first, rather than committing to a full launch. Those constraints matter: an attractive idea that requires a new team may be unsuitable for this decision.
Use a prompt like this, replacing the example details with your own:
Help me prepare a decision for our design studio. We are choosing between monthly support for existing clients, a fixed-price workshop for new clients, and continuing our current project work. Our goal is more predictable work without hiring this quarter or delaying existing projects. First, restate the decision and constraints. Ask for information that could materially change the choice. Do not recommend an option yet.
Read the restatement before continuing. If ChatGPT assumes that growth means maximizing revenue regardless of workload, correct it. The first useful output is a decision your team recognizes, with agreement about what it must protect.
2. Prepare a small, dated evidence pack
Collect the records that bear on the choice. For the studio, these might be completed projects, client requests, delivery capacity and notes from customer conversations. Give each source a short label and include its date or reporting period, so you can trace a claim back to the original.
A simplified fictional pack:
| Source | Illustrative input | What it leaves unanswered |
|---|---|---|
| R1: project and request log, last completed quarter | Four of twelve clients requested small updates after their main project | Whether those clients would pay for monthly support |
| R2: delivery calendar, checked today | Eight delivery days remain uncommitted next month | How much time a new service would consume |
| R3: notes from two recent client conversations | Both clients expressed interest in ongoing help | Acceptable price, scope and willingness to commit |
Share only information permitted in your chosen workspace. Remove identifying details that the analysis does not need, and check your organization’s rules before uploading client documents or confidential figures. If you cannot share a record, prepare an authorized summary or keep that part of the analysis outside ChatGPT.
For repeated planning work, Projects in ChatGPT can organize related chats, files and instructions. It is an optional way to keep context together. A simple document with source labels also works as your planning record; organizing files does not establish that their contents are accurate or current.
Before analysis, check that the sources use compatible dates and definitions. A capacity figure from next month and a sales figure from last year may both be accurate while answering different questions.
3. Separate facts, assumptions and missing information
Ask ChatGPT to sort the evidence before it proposes a strategy. A fact is something supported by the supplied record. An assumption is something you are treating as true for the analysis but have not established. An unknown is information you still need.
In the studio example, R1 supports the statement that four clients asked for updates. It does not establish demand for a subscription. R3 gives the owner a reason to investigate monthly support, but expressions of interest are different from accepting a defined offer at a stated price.
This prompt makes that distinction explicit:
Using only sources R1 to R3, list the facts relevant to our decision, the assumptions we would need to make, and the important unknowns. Give the source label for each fact. Flag conflicting or outdated information. Do not fill gaps with plausible figures. For each unknown, suggest a way we could investigate it and explain whether it could change the decision.
Check the output against the records. If a claim has no source, move it to assumptions or remove it. If two records disagree, ask the person responsible for them to resolve the difference before treating either as settled.
External research may help when the missing information concerns a market or competitor. Where available, ChatGPT’s deep research tools can produce reports with source links. Availability depends on your plan, location and workspace settings. You can also gather relevant sources yourself. In either case, inspect the original material: a competitor’s pricing page may establish an advertised offer, but it does not reveal their profit, customer retention or future strategy.
When an unknown could reverse the choice, gathering evidence becomes the next action. Another round of prompting cannot tell the studio whether its clients will buy an offer they have not seen.
4. Compare realistic options, including continuing as you are
Choose the criteria before asking for a recommendation. For the studio, useful criteria are evidence of demand, delivery effort, effect on existing clients and how much commitment is needed to learn. Apply the same criteria to each option, including continuing the current service.
Use the same questions for each option:
Compare monthly support, a fixed-price workshop and continuing project work against evidence of demand, delivery effort, effect on existing clients and the commitment needed to learn. Use R1 to R3, cite the relevant source labels and distinguish evidence from assumptions. Where the pack cannot answer a question, mark it unknown. Do not treat missing evidence as proof that demand is absent or invent scores.
The evidence part of that comparison would look like this:
| Option | Evidence available in this example | Main question before committing |
|---|---|---|
| Monthly support for existing clients | Repeat requests and two expressions of interest | Will clients pay for a defined scope that the studio can deliver? |
| Fixed-price workshop for new clients | No direct customer evidence in the supplied pack | Who needs it enough to buy, and what would reaching them require? |
| Continue project work | The project log describes the current business | Can current work meet the owner’s goal, or does its variability need addressing? |
Ask ChatGPT to explain the trade-offs behind its comparison. If it assigns scores, require the basis for them. A five-star demand rating created from two interested clients is still an assumption, however tidy the table looks.
For this fictional business, testing support with existing clients is a reasonable first investigation because the pack contains relevant requests and people to speak with. That is a provisional judgment about where to learn next. It is not evidence that retainers will become the studio’s best growth strategy.
5. Challenge the option that looks strongest
Once an option starts to look attractive, use ChatGPT to examine the case against it. Ask about the conditions under which it would fail, what the team may be overlooking and which evidence would change the recommendation.
Research illustrates why task-by-task checking matters. In a randomized experiment involving 758 BCG consultants, researchers compared access to GPT-4 with no AI access. They found improved performance on tasks within the model’s capabilities but worse correctness on a task outside them. The study by Dell’Acqua and colleagues concerns those consultants, tasks and model. It does not establish the accuracy of today’s ChatGPT or predict returns for your business. It does show why useful performance on one task is insufficient grounds to trust the next.
For the studio, the strongest objection may be delivery effort. Clients could value ongoing help but expect much more work than the proposed fee covers. A second concern is timing: support requests might arrive when existing projects need the team’s attention.
Try this prompt after the comparison:
Challenge our provisional choice to test monthly support. Use the evidence pack and distinguish supported concerns from hypothetical ones. Describe a favorable scenario and a downside scenario, stating the assumptions in each. Identify the assumption that would most change the choice, the evidence needed to check it, and a small test before a wider launch. Do not invent probabilities or customer commitments.
Treat the resulting scenarios as conditional descriptions, not forecasts. Check the arithmetic independently in a spreadsheet or calculator. If three hypothetical clients each need three delivery days, that means nine days against the eight available in R2. A persuasive explanation does not remove the capacity gap.
6. Make a human decision and define the test
Bring the comparison and the challenge to the person responsible for the decision. They should consider what the records miss, including commitments to customers and the team’s practical experience. ChatGPT can help draft the decision record once that person has made the choice.
For the studio, a proportionate choice might be to offer a one-month support pilot to two existing clients. That keeps the commitment limited while testing willingness to pay and delivery effort. The following record is illustrative; it is not a reported result.
| Decision field | Fictional studio example |
|---|---|
| Choice | Test monthly support with up to two existing clients before considering a wider launch |
| Reason | R1 shows repeat requests; R3 identifies interest worth investigating |
| Open questions | Will clients accept the price and scope? Can the work meet the studio’s margin requirement? |
| Owner | Studio lead, with delivery estimates checked by the person doing the work |
| Limits | No hiring; at most two delivery days per client; no delay to existing commitments |
| Evidence to collect | Accepted offers, fees, actual delivery and coordination time, direct costs, and client feedback |
| Review | Set a calendar date 30 days after the pilot starts; review earlier if a limit is at risk |
Before making the offers, the owner must define the price, included work and minimum acceptable margin. Write those acceptance conditions down before seeing the results. Two paying clients would establish something about willingness to buy that particular offer; they would not by themselves establish that the service is profitable or ready to scale.
If the main uncertainty is demand, the next step is customer evidence. Our guide to opportunity discovery explores how to move from an attractive idea toward a proposition worth testing.
7. Review the decision and the planning process separately
At the review, compare the actual results with the assumptions you recorded. If clients declined, investigate why. If they bought but delivery exceeded the limit, revise the scope or price before expanding. If the evidence is mixed, record what remains uncertain and decide whether another bounded test is justified.
Also examine whether ChatGPT helped with the planning work. Track preparation and checking time, unsupported claims you had to remove, and useful alternatives the team considered. A commercially successful pilot does not prove that ChatGPT caused the success. Equally, rejecting an unsuitable option may be a useful planning outcome even though it produces no new revenue.
Before closing the exercise, confirm that the decision has traceable evidence, explicit assumptions, alternatives, an accountable owner and a dated review. Reopen it when a significant input changes, such as demand, pricing or available capacity. Keep the source pack and decision record together so the next discussion starts from what the team learned.
Start with the next decision your team needs to make
Choose an upcoming business decision and write its constraints before opening ChatGPT. Gather the few records that could change the answer. That gives you a concrete basis for asking the assistant to organize, compare and challenge your thinking.
If your question is where AI belongs across the company, our separate guide to writing a company AI strategy addresses that broader choice. When evidence gathering and decision preparation become recurring work across teams, they can also form part of a wider AI transformation effort.
Untaylored is an AI Transformation Studio working across AI strategy, transformation and build. To discuss a planning process you want to improve, book an introductory conversation and bring one decision, the information behind it and the point where your team gets stuck.
