Where is the value in advisory services when AI makes analysis cheap?
Adrian E. Urgena · Founder & CEO, Mezisa
Every serious advisory engagement ends at the same moment. The deliverable goes to the people who have to act on it: an investment committee, a lender’s credit officer, a board. Nobody in that room asks how many hours the model took to build. They ask whether they can rely on it.
Answering that question affirmatively is the core value proposition of advisory work, which has traditionally been sold as an indivisible bundle combining both analysis and professional judgment. AI is now separating the two, and that separation will decide where the value and fees are captured.
Two products, sold as one
Analysis is the work of gathering the evidence, building the model and drawing conclusions from it. Analysis can include judgments and recommendations; for example, in transaction advisory a well-built diligence analysis recommends risk-reducing and value-enhancing actions. Advice is something different: qualified professional judgment applied to that analysis. The advisor sharpens the analysis and then stands behind the final output as fit for client action.
Traditional engagements bundled the two because the same firm produced both. Junior teams produced the analysis, senior professionals supplied the judgment, and the client paid for the hours. Most of those billable hours were used to produce the analysis. The bundle obscured a simple fact: clients were paying mostly for production — and buying judgment along with it.
AI is pulling the bundle apart
With AI, the production of analysis is becoming cheap, and it gets cheaper with each model generation. Quality-of-earnings (QoE) work, a staple of mid-market transactions, already shows the spread. A CPA-led boutique publishes a flat-fee range of $15,000 to $30,000 for acquisitions up to $30 million in enterprise value.¹ At the other end, a software vendor publishes a price of $1,000 per project for the QoE analysis itself, or $5,000 with a licensed CPA’s review added.²
The largest firms see the same thing. In March, PwC’s US chief executive said the firm would turn some tax and consulting services into AI tools that clients can use “without a PwC person in the loop,” paid for by annual subscription.³ That is analysis sold as analysis, and priced as a product.
None of this means clients no longer need advice. It means they can now distinguish between the pieces for which they are paying.
What clients still pay a third party to provide
The people who act on advice want both. In a survey of 1,000 senior dealmakers in corporate development, private equity, law, accounting and professional services, conducted by FT Longitude for the deal-platform company Datasite and published in June, 62% said human-only decision-making is no longer defensible in complex transactions. Yet 58% rely on human review and validation to build trust in AI-generated outputs.⁴ They want the analysis AI can produce, and they want a qualified person to stand behind it. A report nobody stands behind is not something on which a board can act.
The companies that sell the analysis draw the same line, and one of them has now drawn it in an SEC filing. Anthropic’s registration statement for its initial public offering, filed in October, states its belief that what it sells is a service rather than a product for product-liability purposes, and relies on warranty disclaimers and limits of liability in its terms.⁵ It does take on one kind of output risk: in some agreements it indemnifies counterparties, without a cap, against claims that its models’ outputs infringe intellectual property. It does not take on whether the output is right. Its usage policy requires human oversight of high-stakes automated decisions, and its product for financial advisors builds the advisor’s review into each workflow. The vendor stands behind the risk that is bounded and legal, and leaves the reliance judgment with the professional. That is the line between analysis and advice, drawn by the seller of the analysis.
What clients will keep paying a third-party advisor to provide is the judgment. It does two jobs, and both matter.
The first is to sharpen the analysis: to see what the data cannot show on its own, to know which anomaly matters and which is noise, and to bring the perspective of having seen the pattern before.
The second is to stand behind the result, so that the people acting on it can rely on a named professional’s view that it is fit for the decision in front of them.
That judgment comes in different strengths. Its minimum form is a review: a qualified professional reads analysis produced by others and writes a memo accepting or rejecting it. Its stronger form works inside the analysis as it is produced, testing and redirecting it before it hardens into a conclusion. With vendors now writing human review into their own products, the nature and quality of review matter more, not less.
The two jobs belong together. A professional cannot credibly stand behind analysis they did not shape, and shaping analysis without standing behind it only produces more analysis.
Where the price goes
Once analysis and advice can be bought separately, the market prices them separately, and the two kinds of offering head in different directions:
Offerings that were really selling analysis are left competing with software, at software prices. Polished deliverables, proprietary templates and large teams do not change that if what the client is paying for is mainly the production of analysis.
Offerings that bundled analysis with judgment keep their demand, especially where a lender, investor or acquirer requires a qualified third party. Their price, though, compresses toward what that judgment is worth, because the analysis inside the bundle no longer justifies an analyst-hour billing of its own.
For advisory firms, the question is uncomfortable but simple. If the analysis in your engagements were priced at what software charges for it, what would remain of the fee, and could you explain what that remainder buys? Firms whose answer is “the judgment, and here is how it improved the answer” are well placed. Firms whose economics depend on billing analyst hours will find the software comparison arriving in their pricing conversations, whether or not they raise it.
For buyers, the useful questions change too. Not how large the team is or how many hours the work took, but who sharpened this analysis, how their judgment changed it, and who stands behind it when it is challenged.
What would challenge this
This is a prediction, and it should be testable. If fixed-fee prices for analysis-only deliverables, such as a quality-of-earnings report or a standalone valuation, hold steady or rise as AI tooling spreads, the argument is suspect.
The bet is the opposite. Analysis was never the advice; it was the work that made the advice possible, and it is about to become inexpensive. What remains, and what will be worth paying for, is the judgment that makes analysis actionable.
About this piece
Drafted with AI assistance from Mezisa’s market research. Every factual claim is sourced below. Reviewed, revised and adopted by Adrian E. Urgena, who stands behind it.
Adrian E. Urgena is founder and CEO of Mezisa Systems. He has over 25 years of operator and business advisory experience across more than 150 engagements and $30B in M&A deal value. He holds a BA in Economics from Stanford and an MBA from the University of Denver.
Sources
Bedrock QoE, “QoE Report Pricing” (accessed November 4, 2026). https://bedrockqoe.com/pricing
Shepi, “shepi DIY vs. shepi DFY vs. Traditional CPA Firm” (updated May 2026; accessed November 4, 2026). https://shepi.ai/compare/ai-qoe-vs-traditional. Archived: https://web.archive.org/web/20261104181641/https://shepi.ai/compare/ai-qoe-vs-traditional
Financial Times interview with Paul Griggs, March 2026, as reported by CPA Practice Advisor, “PwC CEO says partners who resist AI have no place at the firm” (March 25, 2026). https://www.cpapracticeadvisor.com/2026/03/25/pwc-ceo-says-partners-who-resist-ai-have-no-place-at-the-firm/180245/
Datasite, “Most dealmakers say human-only decision-making is no longer defensible in complex deals” (press release, June 24, 2026), reporting a survey of 1,000 senior dealmakers in 27 countries conducted by FT Longitude in March 2026. https://www.datasite.com/en/company/news/most-dealmakers-say-human-only-decision-making-is-no-longer-defensible-in-complex-deals
Anthropic, PBC, Form S-1 registration statement (preliminary prospectus), filed October 26, 2026: p. 50 (service characterization; warranty disclaimers and limitations of liability), p. 93 and p. F-61 (uncapped intellectual-property indemnity), pp. 231–232 (usage policy requiring human oversight of high-stakes automated decisions), p. 203 (advisor review in Claude for Financial Advisors). https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm
Start with a conversation.
Whether you’re weighing a growth, performance or exit decision, or want to compare notes on where advisory is going next, we welcome the conversation.
