In the last eighteen months, the most common pattern I have seen on first calls with prospective SGI clients is the founder who has spent forty to sixty hours with ChatGPT building their commercial case and is now wondering why institutional investors are not engaging with it. The work is, on the surface, polished. The business plan reads well. The financial model is internally consistent. The market sizing has TAM, SAM, and SOM figures. The pitch deck has clean visual logic. And yet conversion through the institutional funnel is flat, and the founder cannot understand why.
Here is the uncomfortable truth that most AI-versus-consultant comparisons soft-pedal: ChatGPT and the current generation of large language models are genuinely useful for parts of the early-stage strategic process. The failure modes are not that AI produces obviously bad work. The failure mode is that AI produces plausible-looking work that contains structural errors no founder catches because the errors are invisible from the founder’s own perspective, and those errors are the ones that kill institutional rounds in the diligence room rather than in the pitch.
This piece compares ChatGPT against a UK startup consulting company honestly. It identifies what AI does well, the three specific failure modes that consistently cost founders rounds, what a consultant adds that AI structurally cannot, and the hybrid workflow that is actually how serious founders are using both tools in 2026.
What ChatGPT Does Well for Early-Stage Strategy
ChatGPT and the current generation of large language models are genuinely useful for several parts of the early-stage strategic process, and pretending otherwise produces poor consulting advice.
Drafting and structuring documents. Given a coherent brief, ChatGPT produces competent first drafts of business plan sections, pitch deck copy, financial model assumptions, market analysis summaries, and supporting documentation. The drafting work that would take a founder ten hours in Word and Excel can take ninety minutes with AI assistance.
Research synthesis at speed. Given a focused query, ChatGPT can synthesise publicly available information on a sector, a competitor, or a regulatory framework faster than a founder can do manually. The output requires verification, but the synthesis step itself is genuinely accelerated.
Editorial and language polish. ChatGPT is reliable at improving readability, fixing grammar, adapting tone for different audiences, and producing variations on copy. For founders who are not natural writers, the editorial assistance is meaningful.
Brainstorming and ideation. As a thinking partner for generating options (alternative pricing strategies, possible market segments to test, framings of the value proposition), ChatGPT is useful because it produces a volume of plausible options quickly. The founder still has to evaluate which options are credible.
Templating and formatting. Standardised structures (business plan headings, financial model layouts, pitch deck flows) are well within ChatGPT’s capability. The structural work that would take a founder several hours to research and assemble is delivered in minutes.
For these five functions, ChatGPT is a productivity multiplier. The founders who use it effectively for these functions free up time and energy for the higher-leverage work that AI cannot do.
The Three Failure Modes That Cost Founders Rounds
ChatGPT fails in three specific aspects of early-stage strategy, and these failures determine whether institutional capital is committed. The structural reason for the failures is consistent: ChatGPT optimises for plausibility rather than for adversarial defensibility, and institutional investors evaluate plans adversarially.
Failure mode one: founder bias amplification. ChatGPT is a language model trained to produce useful, on-brief output. When a founder briefs it with their own assumptions about their market, competitors, unit economics, and differentiation, the model produces an analysis that uses those assumptions as inputs and builds the case on top of them. The model does not have access to the contrarian evidence that would challenge the assumptions; it has access to the founder’s own framing of the assumptions, and it builds plausible reasoning on top of that framing. The result is a business case that confirms what the founder already believed, presented in confident analytical language. A founder who believes their TAM is £4 billion when it is actually £400 million will, with AI assistance, produce a £4 billion TAM analysis that reads convincingly. The institutional investor’s first-round market diligence will identify the gap in twenty minutes.
A Birmingham-based ed-tech founder I worked with had used ChatGPT to develop a sixty-page commercial case before our first call. The case argued that the addressable UK secondary-school market for the product was 4,200 schools, generating £8M of ARR potential at fully loaded penetration. The actual, realistic addressable market, once we worked through the procurement constraints, budget cycles, and decision-making structures in UK state and independent schools, was approximately 280 schools with a realistic revenue potential of £1.4M. The ChatGPT-assisted case had taken the founder’s framing as gospel and amplified it. The institutional investors had identified the gap immediately in their own diligence and declined to engage further.
Failure mode two: absence of consultative pushback. A good consultant disagrees with the founder when the founder is wrong. ChatGPT is structurally optimised to be agreeable. The model will sometimes flag concerns when prompted directly, but it does not autonomously identify and push back on weak reasoning the way a consultant who has worked with 2,000 businesses over 25 years does. The result is that founders using AI as their strategic partner do not get the consultative friction that exposes weak thinking before it reaches the investor diligence room.
The consultative pushback is structurally what most founders are buying when they engage a consultant. The polished deliverable is the visible product; the disagreement during the engagement is the actual product. ChatGPT cannot disagree in the same way because it is not coming from a position of independent commercial judgment; it is coming from a position of helpful response generation.
Failure mode three: no funder calibration. Institutional investors at different stages and across sectors use different evaluation frameworks. A pre-Seed VC investing in deep-tech evaluates a plan differently from a Seed-stage SaaS investor or an angel network specialising in consumer brands. The differences are not cosmetic; they are structural — what the funder weighs, what evidence they require, what red flags they pattern-match against, what TAM construction they accept. A consultant who has placed dozens of plans with specific funders calibrates the documentation to that funder’s framework. ChatGPT produces plans calibrated to a generic average of investor expectations, which means the plans miss the specific calibration needed to close specific rounds.
SGI’s relationships with named VCs (Atomico, Balderton Capital, Index Ventures, Octopus Ventures, Seedcamp, LocalGlobe) and angel networks (SFC Capital, OION, Cambridge Angels) inform exactly this calibration. When I work with a founder targeting Seedcamp specifically, I know what the Seedcamp partners look for and structure the engagement and the deliverables around that. ChatGPT cannot replicate this calibration because it lacks working knowledge of the specific funder’s evaluation patterns.
What a Consultant Adds That AI Cannot
The functions a consultant performs that AI cannot perform structurally are those that determine institutional funding outcomes. There are five.
Adversarial diligence simulation. A consultant who has worked with funded ventures simulates the investor’s diligence questions before they are asked. The work the consultant does in the engagement is closer to investor diligence than to plan drafting; the plan is the visible deliverable, but the actual product is the diligence-stress-tested commercial case. ChatGPT does not simulate adversarial diligence because its training optimises for generating helpful responses rather than for adversarial probing.
Funder-specific calibration. A consultant with named funder relationships calibrates the engagement to the specific framework that the targeted funder applies. The plan that wins Atomico’s attention is differently structured from the plan that wins Octopus Ventures’ attention, and a consultant with both relationships in their working knowledge can tailor accordingly.
Operational diagnosis from observed signals. A consultant in the room with the founder picks up signals about the business and the founding team that no AI workflow detects: how the founder talks about their customers, whether the technical co-founder has commercial instinct, where the founder is over-claiming and where they are under-claiming. The diagnostic informs the consulting work in ways that document-based AI workflows structurally miss.
Network access. A consultant with working relationships in the UK investor ecosystem can warm-introduce founders to specific VCs and angels where appropriate, advise on which firms are currently active in the sector, and provide informal market intelligence on what is and is not closing. AI workflows produce documents; consultants produce documents and access.
Implementation accountability. A consulting engagement that runs through to funding close maintains accountability for the work, translating into the outcome. The consultant is engaged for the period during which the plan is being tested with investors and can iterate as feedback comes in. AI workflows end when the document is produced.
The Cambridge deep-tech engagement (Planetary Processing) illustrates the difference. Twelve weeks of customer discovery, financial modelling, business plan development, and funder calibration culminated in a round closed with Blue Wire Capital, Cambridge Enterprise, and Creator Fund. The work that closed the round was not the document itself; it was the iterative, consultative work of stress-testing the commercial case against the specific funders’ frameworks before the meetings. ChatGPT could have produced the document. It could not have produced the funded outcome.
The Hybrid Workflow That Actually Works
The pattern I see among the most effective founders in 2026 is hybrid: using ChatGPT aggressively for the work it excels at, with a consultant engaged for the work AI is structurally incapable of doing.
Stage one: ChatGPT for drafting and research synthesis. The founder uses AI to produce competent first drafts of business plan sections, market research summaries, financial model assumptions, and supporting documentation. This compresses the documentation work from forty to fifty hours into perhaps fifteen, freeing time and energy for the higher-leverage work.
Stage two: consultant engagement for diagnostic and adversarial review. The consultant takes the AI-drafted material as input and applies the work AI cannot do: adversarial diligence simulation, founder-bias correction, funder-specific calibration, operational diagnostics from interview signals, and consultative pushback that exposes weak reasoning. The output is a commercial case that has been stress-tested against the institutional standard.
Stage three: iterative refinement. The consultant and the founder iterate on the documentation through the early investor meetings, taking feedback from each meeting and incorporating it into the next version. This iterative work is structural to closing institutional rounds, and it is not work that AI can perform.
The cost economics of the hybrid workflow are typically favourable. A founder who uses ChatGPT for 15 hours of drafting and £2,500 in consulting for diagnostic and adversarial work spends less total time and money than a founder doing 50 hours of solo work with ChatGPT, plus a third-party consultant retained at the last minute to fix structural problems that the diligence room has surfaced.
Common Mistakes in the AI-Versus-Consultant Decision
Three patterns reliably produce poor outcomes.
Treating AI as a complete substitute. The most common pattern is the founder who concludes that ChatGPT’s existence makes paid consulting unnecessary. The documentation comes out polished, the founder feels productive, and the structural problems are invisible until the diligence room exposes them. The Birmingham ed-tech case is one of dozens I have seen.
Treating AI as worthless. The reciprocal mistake is the founder who insists that AI is not useful and continues to spend 40 hours producing documentation manually, which could have been drafted in 15 with AI assistance. The time saved is meaningful in itself; refusing to use AI on principle costs the founder weeks of compounding work.
Outsourcing the strategic thinking. Whether to AI or to a consultant, the founder who outsources the strategic thinking has fundamentally misunderstood the value of the engagement. AI accelerates drafting; consultants accelerate diagnostic and calibration. Neither replaces the founder’s ownership of the strategic position. Founders who use either tool as a substitute for thinking are buying the documentation product without the thinking product; documentation alone does not close rounds.
Decision Framework
The framework I use with founders is simple.
If your need is documentation production and editorial assistance: ChatGPT, used carefully with verification of every claim and source, will significantly accelerate the work.
If your need is funder-calibrated commercial case development for institutional capital: ChatGPT alone is structurally inadequate, and the hybrid workflow (AI for drafting, consultant for diagnostic and calibration) is the right approach.
If your need is adversarial diligence simulation against specific funders: a consultant with named funder relationships is the only credible source. AI cannot calibrate to specific institutional patterns the way a consultant who has placed dozens of plans with the named firms can.
If you are raising less than £50,000 of debt funding (typically a Start Up Loans application): ChatGPT plus a defined-scope business plan service is often the right combination. The full consulting engagement is calibrated for larger raises with institutional capital.
Conclusion
The principle underneath this entire piece is that AI and consulting are not substitutes in the early-stage strategic process. They are complements that address different parts of the work, and the founder who understands which part each tool serves will produce better work, in less time, for less total cost than the founder who treats one as a replacement for the other.
ChatGPT is a powerful productivity tool for drafting, research synthesis, and editorial work that consumes a meaningful proportion of the early-stage strategic process. A consultant with funder relationships and adversarial diligence experience is the structural answer to the work that AI cannot do: bias correction, funder calibration, operational diagnostics, network access, and implementation accountability through to funding close.
The right approach is not to choose between them. The right approach is to use each for what it does well, and to be honest about which part of the process is which.
Next Step: Free Startup Assessment
If you have built early-stage strategic material with ChatGPT and want a consultative review before going into institutional meetings, the SGI startup assessment is a 45-minute call. We will review the work, identify the structural gaps that the diligence room will probe, and recommend whether a defined-scope engagement is appropriate to close the gaps before the round.
Visit our Startup Consultants service page for methodology and pricing.
FAQ
Can ChatGPT write a business plan for institutional investors? ChatGPT can produce a competent first draft of a business plan structure. It cannot produce a plan that closes institutional capital, because the work that closes institutional rounds is the adversarial diligence simulation, funder calibration, and bias correction that AI structurally cannot perform. The hybrid workflow (AI for drafting, consultant for diagnostic and calibration) is the right approach for institutional fundraising.
Will AI replace startup consultants? For the drafting and research synthesis components of consulting, AI has already significantly compressed the work. For the diagnostic, calibration, network access, and accountability components, AI is not a substitute. The market is shifting toward hybrid workflows where AI handles drafting and consultants handle the higher-leverage work; the consulting market is not being replaced, it is being repositioned around the work AI cannot do.
Should I use ChatGPT to prepare for a YC or Seedcamp application? ChatGPT is useful for drafting the application content and synthesising research on the programmes. It is structurally inadequate for calibrating specific programme partners’ decision-making, which determines whether the application converts. For founders serious about Seedcamp or YC acceptance, a consultant with knowledge of the specific programme’s selection patterns is materially more valuable than AI on the final application work.
How much does it cost to combine ChatGPT and a startup consultant? The hybrid workflow typically costs £800 to £3,000 in consulting fees, plus the founder’s time spent using AI for drafting. The combined cost is materially less than retaining a consultant for full drafting and diagnostic work, and materially less than the opportunity cost of a failed institutional round caused by AI-only documentation that did not stand up in diligence.
Is AI better for some sectors than others when developing startup strategy? ChatGPT performs better in sectors with abundant public documentation (consumer SaaS, e-commerce, generic B2B services) and worse in sectors where the operating constraints are concentrated in non-public knowledge (regulated financial services, deep-tech IP commercialisation, specialist B2B niches). The structural reason is the training data; AI knows what it has been trained on, and specialist sectors are under-represented.
Can I trust ChatGPT’s financial projections? No. ChatGPT can produce financial models that are internally consistent, but its projections are built on the assumptions you provide; if the assumptions are wrong, the model amplifies the error rather than correcting it. Financial projections that will be evaluated by institutional investors should be prepared or stress-tested by a financial professional with sector-benchmarking experience.
References
- Princeton GEO Study (KDD 2024). Generative Engine Optimisation. Aggarwal et al. https://arxiv.org/
- British Business Bank. Small Business Finance Markets Report 2024. https://www.british-business-bank.co.uk/
- ONS. Business Demography UK 2023. Office for National Statistics. https://www.ons.gov.uk/
- Beauhurst. The Deal: UK Equity Investment 2024. https://www.beauhurst.com/
- British Venture Capital Association. BVCA Performance Measurement Survey 2024. https://www.bvca.co.uk/
Kurt Graver is the founder and CEO of SGI Consultants, a business consultancy that has helped over 2,000 entrepreneurs establish successful startups using systematic business development methodologies. An accountant with an MBA and 25 years of commerce and consultancy experience, Kurt specialises in strategic planning, market analysis, and sustainable business growth