25 ChatGPT Prompts for Consultants: Strategy and Decks (2026)

Most lists of ChatGPT prompts for consultants are written by people who think consulting is fancy slide-making. They give you "create a SWOT analysis" prompts and pretend the tool is doing the work that clients actually pay for.

The work clients pay for is the thinking: framing the problem the client did not know they had, separating the load-bearing facts from the noise, and writing the page that the partner can defend in front of the executive committee on Thursday. ChatGPT can help around the edges of that work. It cannot do the work itself, and you should not let it pretend to.

This list covers prompts for the consulting tasks the model is genuinely useful for: brief synthesis, framework drafting, deck outlining, executive comms, and the operational work that fills the calendar between client moments. Every prompt has the role, task, constraints, and output shape pinned. The bracketed inputs are what you fill in with your actual engagement.

How to use these ChatGPT prompts for consultants

A few things to know before you open a chat.

ChatGPT is good at the structured-writing tasks: turning interview notes into themes, drafting an executive summary from your own bullets, generating five candidate framings for a problem, writing the meeting recap nobody else has time to write. It is bad at the work that requires being inside the engagement: judging which client stakeholders matter, calling the political risk of a recommendation, knowing what the partner will and will not accept on Friday. Treat it as a senior associate who has read the deck once and is happy to draft, not as a strategist who has been in the room.

Do not paste client data into the free tier. Engagement notes, client names, financial figures, and interview transcripts are confidential by default, and the free tier may use your inputs for model training. Anonymise before pasting (rename the client, scrub the figures by a constant factor, strip the named stakeholders), or use a paid plan with chat history disabled, or use your firm's enterprise AI tool. For any work that touches MNPI (material non-public information) or PII, the answer is always the enterprise tool, never the free tier.

The prompts below all assume real, specific inputs. Vague inputs produce the kind of "frameworks" that get a junior consultant in trouble with a partner. Replace the bracketed examples with your actual engagement details.

Engagement scoping and brief synthesis

Five prompts for the work that happens in the first two weeks of an engagement.

1. Synthesize interview notes into themes

Prompt: "Below are notes from [N] stakeholder interviews on [engagement topic]. Cluster the observations into 4-6 themes. For each theme: a 1-line label, the 2-3 strongest pieces of evidence with the speaker initials only, one tension or disagreement across speakers, and a 1-line note on what this theme implies for the recommendation. Constraint: only use facts in the notes. Do not invent quotes. If a theme has only one speaker behind it, label it 'single-source' so I can flag it for the partner. Notes: [paste]."

The "do not invent quotes" instruction is load-bearing. The model will paraphrase plausible-sounding interview quotes if you do not pin this.

Prompt: "I am scoping a 6-week engagement for [client industry and size] on [engagement topic]. The client's stated objective is [paste]. The unstated objective is likely [your read]. Write a 1-page scope document covering: (1) the problem in 25 words, (2) what is in scope (5 bullets with workstream names), (3) what is explicitly out of scope (3 bullets, so a junior can defend the boundary), (4) the deliverables and dates, (5) the named stakeholders we will need from the client side. Constraint: under 400 words. No filler. No 'thought leadership'."

The unstated-objective line is the difference between a scope that gets signed and one that gets renegotiated week three.

3. Generate 5 candidate problem framings

Prompt: "The client described the problem as: [paste their words]. Generate 5 alternative framings of the same problem. For each: a 1-line restatement, the assumption it changes, and the type of recommendation it would lead to. Constraint: at least 2 of the framings should be ones the client probably does not want to hear. No 'opportunity' euphemisms. Output as a numbered list."

The "framings the client probably does not want to hear" line is what makes this prompt actually useful. Without it, the model gives you five polite restatements that all lead to the same generic recommendation.

4. Build a stakeholder map

Prompt: "I am working with these [N] named client stakeholders on [engagement]: [list with role and what you have learned about each from any interviews so far]. Map them on two dimensions: influence (1-5) and orientation toward the recommendation (skeptic / neutral / sponsor / champion). For each: the placement, the specific reason for the placement, the one thing they need to hear from us, the one thing they should not hear yet. Output as a markdown table with columns: Name | Role | Influence | Orientation | Needs to hear | Should not yet hear."

This is the prompt that should run on day three of every engagement, not day twenty.

5. Draft a kickoff agenda

Prompt: "Draft a 60-minute kickoff agenda for [engagement]. Attendees: [list, roles only]. Goals of the meeting: (1) align on the problem statement, (2) confirm scope and deliverables, (3) agree the working cadence, (4) identify the named client decision-makers. Output: time-blocked agenda with 5-10 minute slots, named owner per slot, a 1-line desired outcome for each. Constraint: leave 10 minutes unallocated at the end for questions. No 'icebreaker.' No 'introductions' beyond 2 minutes."

The "no icebreaker" guardrail is necessary because the model defaults to including one. Most consulting clients hate icebreakers and have heard the same partner's introduction six times.

Framework drafting and analysis structure

Five prompts for the analytic work that fills the middle of an engagement.

6. Draft a hypothesis tree

Prompt: "Build a hypothesis tree for [the recommendation we are testing, e.g., 'X client should enter the European market in the next 18 months']. Levels: (1) the main hypothesis, (2) 3-5 sub-hypotheses that must each be true for the main to hold, (3) under each sub-hypothesis, 2-3 testable questions and the data we would need to answer them. For each terminal question: the data source we would actually use (public, client-provided, primary research) and the estimated effort in days. Output as nested bullets."

This is the structure clients pay for. The model produces it cleanly when given a clear main hypothesis.

7. Apply a named framework to a specific case

Prompt: "Apply [framework, e.g., 'Porter's Five Forces'] to [the specific industry and company context]. For each element of the framework: 2-3 facts specific to this case (not generic industry observations), one tension between the elements that is not always covered in the textbook version, the 'so what' implication for the client. Constraint: 'so what' must be a decision, not an observation. If the framework's element does not fit this case, say so instead of forcing it. Output: 1 section per element with bold sub-headings."

The "say so instead of forcing it" instruction prevents the model from generating the kind of forced-fit framework analysis that partners crossed out in red 15 years ago.

8. Write an executive summary of a recommendation

Prompt: "Write a 1-page executive summary of a strategy recommendation. Client: [1-line description]. The recommendation in 1 sentence: [paste]. The 3 strongest reasons it is right: [bullet list]. The 2 biggest risks: [bullet list]. The recommended next step in 1 sentence: [paste]. Audience: a CEO who has 90 seconds. Structure: (1) the recommendation as a single declarative sentence, (2) a 3-sentence paragraph that names the data behind it, (3) a 2-sentence paragraph that names the risks honestly, (4) a single sentence asking for the specific decision we need. Constraint: under 250 words. No 'we recommend.' No 'in summary.' No buzzwords."

This is the prompt I tested for this post. The verbatim ChatGPT response and editor's notes are below.

What this prompt actually returns (tested 2026-05-23)

We ran Prompt 8 on free ChatGPT with a concrete brief baked in: [client: '$1.2B mid-market specialty coffee retailer, 340 stores Midwest and Southeast, 6% YoY growth through 2025, same-store sales flattening in Q1 2026'; recommendation: 'Exit the standalone drive-thru store format over the next 24 months and reinvest the freed capital into a smaller-footprint urban kiosk format anchored on grab-and-go cold beverages'; 3 reasons (drive-thru same-store sales -11% YoY in Q1 2026 vs +4% company average; kiosk pilot 38% gross margin vs 24% drive-thru fleet; 7am-10am and 2pm-5pm urban worker dayparts +14% YoY vs -3% suburban commute daypart); 2 risks ($18M lease-break and remodel costs on 47 stores in 2026; kiosk format only tested in 3 markets); next step: 'Approve a phased 24-month transition starting with the bottom-quartile 12 drive-thru stores by Q3 2026, with a stage-gate decision in Q1 2027']. Reproduced verbatim:

Exit the standalone drive-thru store format over the next 24 months and reinvest the freed capital into a smaller-footprint urban kiosk format centered on grab-and-go cold beverages.

Drive-thru stores have become the company's clear underperformer, with same-store sales declining 11% YoY in Q1 2026 against a +4% company average despite overall revenue growth through 2025. In parallel, the kiosk pilot operated for eight months across three test markets and delivered 38% gross margin versus 24% for the drive-thru fleet while generating similar revenue per square foot. Consumer traffic patterns are also shifting away from the suburban morning commute toward urban worker demand, with the 7am–10am and 2pm–5pm dayparts growing 14% YoY in pilot regions while the suburban commute daypart tied to drive-thru usage declined 3%.

The transition will create a near-term earnings impact, including approximately $18M in lease-break and remodel costs tied to closing 47 drive-thru stores in 2026 before kiosk economics scale. The kiosk format has also only been tested in three markets, creating execution risk if operational performance does not generalize across a broader fleet rollout.

Approve a phased 24-month transition beginning with the bottom-quartile 12 drive-thru stores by Q3 2026, with a stage-gate decision in Q1 2027 tied to early kiosk fleet performance.

Editor's notes. The model held the 4-paragraph structure exactly as the brief asked (1 sentence + 3-sentence data paragraph + 2-sentence risk paragraph + 1-sentence ask) and landed at ~223 words, comfortably under the 250-word ceiling. The banned phrases held cleanly: no "we recommend," no "in summary." All ten numbers from the brief survived intact (11%, 4%, 38%, 24%, 14%, 3%, $18M, 47 stores, 12 stores, Q3 2026, Q1 2027), which is the load-bearing requirement for an executive summary. Three things to fix before sending to the partner: (1) the recommendation sentence swapped the brief's "anchored on grab-and-go cold beverages" for "centered on grab-and-go cold beverages" which is a minor language drift that matters because the partner specifically chose "anchored" in the prior draft; (2) the data paragraph leads with "Drive-thru stores have become the company's clear underperformer" which is editorial framing the brief did not authorize and which a CFO might push back on as overreach (the data shows the format is underperforming the company average for the most recent quarter; calling it "the company's clear underperformer" implies a structural multi-quarter pattern the brief did not establish); (3) the closing single sentence is the strongest paragraph and the one a real CEO will quote back, which means the recommendation ask is doing the work that the deck title should do, suggesting the partner should consider promoting that sentence to the cover slide. Also worth flagging the model's "before kiosk economics scale" phrasing in the risk paragraph is slightly elevated jargon for a CEO-targeted summary and a one-pass edit to "before the new kiosk fleet generates offsetting margin" would land more cleanly. Net: shippable as a 90% first draft after a 5-minute partner edit pass.

9. Generate 8 sensitivity-test questions

Prompt: "Our model says [paste the headline finding, e.g., 'entering Germany next year is NPV-positive at $42M over 5 years assuming 8% terminal growth']. Generate 8 questions a skeptical partner or client would ask to stress-test the finding. For each: the question, the specific assumption it pressures, the data we would need to defend it. Rank by how badly the finding would degrade if the question lands. Do not assume the finding is correct."

The "do not assume the finding is correct" line is what stops the model from producing the friendly version of these questions instead of the hard version.

10. Critique a draft analysis section

Prompt: "Below is a draft analysis section from a deck. Read it like a skeptical partner. Flag the top 5 issues, ranked: (1) factual claims that need a source, (2) any 'analysis' that is actually assertion, (3) anywhere a chart's label or axis would mislead a reader, (4) any chain of logic that skips a step, (5) any place where the conclusion sentence overstates the data. For each: name the issue, point to the specific phrase, recommend the smallest fix. Section: [paste]. Do not rewrite. Output as a numbered list."

The partner-skeptic framing produces sharper feedback than the model's default "this is well-structured, here are a few thoughts."

Deck outlining and slide drafting

Five prompts for the work that fills the back half of an engagement.

11. Outline a 25-slide steering committee deck

Prompt: "Outline a 25-slide steering committee deck for [engagement topic] at week [X] of [Y]. Audience: [steering committee composition]. The 3 decisions we are asking the committee to make: [list]. Outline: cover, exec summary, 3 evidence sections (each with intro slide + 4-6 evidence slides + summary slide), recommendation, risks, decisions requested, next steps, appendix. For each non-section slide: a 1-line title that states the slide's point, not its topic. Constraint: no slide title ending in 'overview' or 'analysis'."

The "title states the point not the topic" instruction is what separates a deck a steering committee can use from a deck that needs the deck author present to translate.

12. Draft an action-title slide

Prompt: "Draft an action-title slide on [the specific slide topic]. Slide title in 12 words or fewer that states the conclusion, not the topic. Three supporting bullets, each starting with a verb or a number, each under 15 words, each citing the source in parentheses. One single chart suggestion that proves the title in one image, named with the chart type and the two axes. Output: Title, Bullets (numbered 1-3), Chart suggestion. Constraints: no 'analysis showed.' No 'in summary.' Verb-first or number-first only."

The verb-or-number-first cap is what makes the bullets read like a partner wrote them.

13. Generate 6 candidate slide titles for the same data

Prompt: "I have this slide draft: [paste current title and 3 bullets]. Generate 6 alternative action-titles for the same data, varying the angle: (1) the implication for the client's biggest business risk, (2) the implication for the recommended action, (3) a contrarian framing, (4) the number that most matters, (5) the comparison to a peer or benchmark, (6) the strongest version of the original. Each under 12 words. No questions. No 'analysis.'"

This is the prompt to run when the partner says "the title does not pop yet" and you need to ship five alternatives by 5 PM.

14. Write speaker notes for a deck

Prompt: "Below is a slide title and 3 bullets: [paste]. Audience: [describe]. Write speaker notes a partner could read verbatim: 2-3 short paragraphs totalling 80-120 words. The first sentence sets up the slide in plain English. The middle gives the 2 strongest pieces of evidence with the source. The last sentence signals the decision the partner wants the audience to make. No 'as you can see.' No 'this slide shows.' Output the notes only."

Saves the post-meeting "what was on slide 14 again" recovery cycle.

15. Build a one-page client take-away

Prompt: "Build a 1-page take-away the client can share internally after our review. Audience: people who were not in the room but will hear about our work. The recommendation: [paste]. The 3 most important pieces of supporting evidence: [bullet list]. The 2 things they need to do next: [paste]. Structure: title (the recommendation as a declarative sentence), 3 sub-sections (each with a bold sub-head, a 2-sentence body, and one specific number), a closing 'what we are asking you to do this month' section. Constraint: under 350 words. No firm jargon. No 'we are excited.'"

Most engagements ship a deck. Few ship a 1-page take-away. The clients that get a take-away spread the recommendation further inside their organisation.

Client communications and meeting comms

Five prompts for the work that lives in email and calendar.

16. Draft a status email a client will actually read

Prompt: "Draft a weekly status email to the client sponsor on [engagement]. This week we accomplished: [bullets, with numbers where you have them]. Risks or open issues: [bullets]. Next week: [bullets]. Decisions needed from the sponsor: [list, in 1 line each]. Structure: (1) one-sentence opener that names the headline of the week, (2) a 5-bullet 'progress' section, (3) a 'decisions needed' section that asks the question in one line each, (4) a 1-sentence sign-off. Constraint: under 200 words. No 'I hope this email finds you well.' No 'as discussed.'"

The "decisions needed" section is the part most consultants leave out and then complain about the sponsor not deciding fast enough.

17. Pre-meeting brief for a partner

Prompt: "I am briefing my partner before a client meeting. Client: [name]. Meeting purpose: [1-line]. Attendees from client side: [list with role]. Where we are in the engagement: [1-2 lines]. The 3 things the partner needs to know that they may not remember: [bullets]. The 1 awkward moment we should plan for: [paste]. Draft a 1-page brief the partner can read in 90 seconds. Structure: (1) headline, (2) what changed this week, (3) the 3 critical reminders, (4) the awkward moment with a recommended position, (5) the single ask the partner has of the client. No firm jargon."

The "single ask" framing forces clarity. Most pre-meeting briefs hedge the ask, and the partner walks in unsure what to push for.

18. Reply to a difficult client email

Prompt: "A client just sent the following email pushing back on a part of our recommendation: [paste, anonymized]. Our position is [paste]. The political reason their email exists is likely [your read]. Draft a 3-paragraph reply that (1) acknowledges the specific concern in their words, (2) explains where we agree and where we still disagree with the data behind it, (3) proposes a concrete next step that lets us both save face. Tone: collegial, factual, not defensive. Do not use the word 'actually.' Do not soften the disagreement into agreement."

The "do not soften" line is the difference between a reply that resolves the issue and a reply that lets the issue grow.

19. Draft a follow-up after a tough meeting

Prompt: "We had a tough client meeting this morning. The 3 things that went wrong: [bullets]. The 1 thing that went well: [paste]. The senior client we need to bring along is [name and role]. Draft a 1-paragraph follow-up email I would send to them today that: (1) names the most important thing we heard, (2) commits to one specific change in how we work, (3) asks for one piece of input from them by [date]. Constraint: under 90 words. No apology theatre. No 'I hear you.'"

Sending this email same-day matters more than the wording. The model removes the friction so you actually send it.

20. Generate 3 candidate openers for a sensitive recommendation

Prompt: "We need to deliver a recommendation the client will not love: [paste recommendation]. The reason it is uncomfortable: [paste]. Generate 3 candidate first sentences for the slide where we land it. Each must (1) name the recommendation in the first 10 words, (2) signal we have respected the client's prior position, (3) avoid passive voice. The 3 should vary across angles: (a) cost-of-not-doing-it, (b) competitive risk, (c) what their best customer would tell them. Constraint: under 25 words each. No 'we believe.' No 'we recommend.'"

Three candidates is the right number. With one option the partner says "is that the best framing"; with three the partner picks.

Practice and admin work

Five prompts for the work nobody pays for directly but everyone has to do.

21. Generate a 90-day plan for a new senior hire

Prompt: "Draft a 90-day plan for a new [role, e.g., 'senior manager'] joining my consulting practice on [date]. Practice area: [describe]. The 3 things they need to know about the team's current strengths and weaknesses: [bullets]. Structure: (1) 30-day priorities (learn the practice, meet the team, get on 2 engagements as observer), (2) 60-day priorities (lead a workstream, write the first internal POV, surface one process improvement), (3) 90-day priorities (own a client relationship, deliver one external talk, propose the next 90-day goal). For each: the deliverable, the named owner of feedback, the success criterion. Output as a markdown table."

The named owner of feedback is what makes the plan real instead of aspirational.

22. Draft a case-study write-up of a finished engagement

Prompt: "Draft a 1-page case study of a finished engagement. Client (anonymised): [describe in 1 line]. Engagement type: [paste]. The client's problem: [paste]. What we did, in 5 bullets: [bullets]. The outcome with the specific number: [paste]. The most surprising thing we learned: [paste]. Structure: (1) headline that names the outcome with the number, (2) 'the situation' (3 sentences), (3) 'what we did' (5 bullets), (4) 'the outcome' (3 sentences with the number repeated), (5) 'what was unusual about this case' (2 sentences). Audience: prospective clients on our website. No firm puffery."

This is the case study format that gets read. The 'what was unusual' section is what builds credibility, because it shows you noticed.

23. Write a 5-question internal research brief

Prompt: "I need a 1-page research brief to scope a 1-week internal research project. Topic: [paste]. Why we care: [1-line]. Audience for the output: [describe]. Structure: (1) the 1-line question, (2) 5 sub-questions whose answers would together answer the main, (3) the data sources we would use for each, (4) the format of the output (memo, deck, table), (5) the success criterion (what we will be able to do with the answer that we cannot do now). Constraint: under 350 words."

Use this for internal POV-building, not client work. Scoping the research saves a week of someone reading the wrong sources.

24. Draft a recruiting message to a passive candidate

Prompt: "Draft a LinkedIn message to a passive candidate for [role]. Candidate: [name, current title, current firm]. The specific reason I am reaching out to them (not generic): [paste]. The role we are filling, in 1 line: [paste]. What is different about our firm for this role: [paste]. Constraint: under 90 words, opens with a specific reference to one piece of their public work (I will paste it in), does not say 'great profile,' ends with an open question. Reference content: [paste 1-line description]. Output the message only."

Generic recruiting DMs are ignored at higher rates than ever. The specific-reference instruction is what flips the response rate.

Prompt: "I want to pitch the firm on writing an internal POV (point of view) on [topic]. Why now: [1-line]. The contrarian take I want to argue: [paste]. The 3 firms or clients I think would care: [list]. Draft a 1-page pitch covering: (1) the question, (2) the contrarian thesis in one sentence, (3) the 3 strongest pieces of evidence I would use, (4) the audience and the format (essay, deck, white paper), (5) the time I would need (in person-weeks). Constraint: under 350 words. The thesis must be contrarian, not just balanced. Output as 5 labeled sections."

The contrarian-thesis cap is what gets the POV read after publication. A balanced POV says nothing the audience did not already know.

Tips for getting better results

Three things that matter more than any specific prompt.

Anonymise inputs before pasting. Rename the client. Strip the named stakeholders. Scale the financials. The information you need ChatGPT to work with is the structure, not the identifying detail. Anonymising costs you 30 seconds and protects an obligation you accepted when you signed the engagement contract.

Pin the unstated objective. Every consulting engagement has a stated objective on paper and a different one in the partner's head. Naming the unstated objective in your prompts is what separates outputs that get used from outputs that get politely thanked.

Treat the first answer as the safe answer. ChatGPT's first version is always the safe, balanced, vendor-neutral version. The useful version comes from the follow-up: "make the recommendation sharper," "give me the contrarian framing," "what is the version a senior partner would push back on." Budget for the second cycle.

FAQ

Is ChatGPT useful enough to draft a deck for a real client engagement?

It is useful enough to draft component slides (action titles, speaker notes, slide outlines) and to critique a deck you have already drafted. It is not useful enough to draft an end-to-end client deck without heavy partner editing. The model can produce a deck-shaped output. It cannot produce a deck that reads as if a senior consultant wrote it, because the load-bearing work (the framing, the so-what, the partner instinct about which slide leads) lives in the relationship not the prompt.

Should I paste client data into ChatGPT?

Not into the free tier under any circumstances. Engagement data is confidential by default and the free tier may use your inputs for model training. Use a paid plan with chat history disabled, your firm's enterprise AI tool, or anonymise the inputs aggressively (rename client, scrub figures by a constant factor, strip named stakeholders) before pasting into any third-party tool. For MNPI or PII the answer is always the firm-approved tool, never the free tier.

Can ChatGPT replace a junior associate on an engagement?

No, and the framing misses the actual question. ChatGPT can take 30 minutes off a 90-minute task for a junior who already knows what to ask. It cannot replace the junior's apprenticeship value (learning to read a room, learning what a partner will push back on, learning to defend a chart to a CFO). Firms that try to skip the apprenticeship to "save associate hours" end up with senior consultants who never learned the basics. Use the model to speed up the junior's work, not to skip the junior.

What is the single most useful prompt for a junior consultant?

Prompt 1: synthesize interview notes into themes. Interview synthesis is the most common first-week task on every engagement and the bottleneck is structure, not creativity. Use the prompt for the scaffold, then your own judgment on which themes are load-bearing.

How do I tell whether a ChatGPT-drafted slide title is good?

Read it without the context of the slide. If it could be the title of a slide on a different deck about a different topic, it is not specific enough. Good action titles are useless out of context, which means they are doing real work in context.

What to do next

Pick one slide in your current deck where the title is not yet right. Run Prompt 13 (6 candidate slide titles for the same data) on it. Notice which of the 6 you would actually pick. The exercise will sharpen what "right" means for the rest of the deck.

If you run a team, share Prompt 1 (interview note synthesis) as the recommended first-week tool. The aggregate quality improvement on early-engagement note synthesis is larger than any single prompt's individual value.

Send one client status email this week using Prompt 16. Check whether the sponsor replies faster than usual. That comparison is the only test that matters.