Your $2M boutique strategy firm just sent a client a 47-page competitive-landscape deck. According to the L40 SaaS Exit Playbook, buyers now ask one question before all others: could the target's own customers rebuild a good-enough version with off-the-shelf AI? A week later, the client's operations manager built a 90% equivalent version in three hours using Claude and a public dataset. The client didn't call back for the execution phase. According to Harvard Business Review, AI is fundamentally reshaping consulting by automating research, modeling, and analysis—the core work of junior consultants that used to create entry points to bigger engagements.[^hbr] If your deliverable is a slide deck or a strategy document, generative AI can produce the raw 80%. The consulting firms that survive the next three years have already passed the "Can your client build it?" test. The ones that haven't won't.
The Math
Consulting multiples haven't decoupled from this reality yet, but they will. The median EBITDA multiple for generic professional services was 3.9x in 2023-2024. Specialized, defensible firms trade at 9.7x to 15.2x.[^valuation] That gap isn't accident. It's measurement of moat. The thin-moat firms:the ones doing commodity work at commodity prices—are getting compressed. Margin pressure is here. Margin collapse is coming. The client's use is "Why hire you when Claude costs $20/month and runs on our procurement system?"
I worked with an Innovation Scout at Hartford Steam Boiler. One of about 15 in a 55,000-person organization. Her job: find the failure modes nobody else was watching. She'd spend three weeks building muscle on a particular risk:say, electrical stress in copper coils:that the standard audits missed. The business paid premium rates and renewed annually. Why? Because nobody else in North America had that specific, earned competence. The client could not replicate it. That's the test.
The difference between a thin-moat firm and a moat-bearing firm is whether the client can answer "yes, good-enough, in-house" to three questions:
Can they replicate your research? If you're pulling public data, cleaning it, and summarizing it, the answer is yes. An AI agent can do that in hours. If you're mining client databases, industry correspondence, or proprietary survey results, the answer is no.
Can they replicate your thinking? If your thinking is a published framework applied by competence, the answer is maybe:but your edge is speed and depth, not novelty. If your thinking is embedded in vertical experience the client doesn't have, the answer is no.
Can they staff it internally? If it takes a partner's judgment and a senior analyst, the client doesn't have $400K sitting on the bench. If it takes their CFO and two people from product, they do:and they will.
The firms passing this test own at least two of these:
1. Vertical depth. You know the industry cold:the regulatory vectors, the talent markets, the cash-flow seasonality, the technology install base. The client knows their business. You know their business and the five other versions of their business they haven't seen yet.
2. Proprietary methodology. Your framework is yours. Not a renamed McKinsey thing or a repackaged Gartner matrix. You built it from work, you've stress-tested it, and it carries your IP.
3. Irreplaceable data. You have survey data the client cannot run. Benchmarks from your closed network. Patterns from 200 implementations they've never seen. Raw material they cannot access.
4. System-of-record status. You become infrastructure. The client doesn't hire you for a project. They hire you for quarterly monitoring, annual optimization, or real-time sensing. You're watching the engine room. You get callbacks because the alternative is flying blind.
The consulting firms that will not survive are the ones executing strategy work, marketing work, IT optimization, operational blueprints:the kinds of engagements where a competent client team plus Claude can get to 75% of the value in one sprint.
The 90-Day Bottleneck Audit
Map your service lines. List every service offering you have that's $50K or larger. For each one, run this quick audit:
Bottleneck 1: Research and Source Work. How much of the deliverable is original source work:interviews, proprietary data collection, or access to networks the client doesn't have? If it's less than 30% of the effort, flag it. If it's zero, plan to exit. An AI can do generic research. It cannot build your Rolodex.
Bottleneck 2: Methodology and Judgment. How much of your value is applying a branded, repeatable method that the client cannot easily replicate? Is the method yours, or is it an industry standard with a prettier name? Can the client hire someone competent and run the method in-house? If yes, your moat is time and coordination only. That is not a moat.
Bottleneck 3: Execution Bridge. Do you move from analysis to implementation? Or do you hand off a report and disappear? The firms that own the implementation phase own the relationship. The client cannot question the analysis if they've already bet capital on it. If you're pure strategy delivery, you're a consultant. If you're advising on execution and staying for the dig-in, you're a partner.
Bottleneck 4: Ongoing Obligation. Do you get repeat work from the same client, same problem, or same vertical? Or is it one-off projectwork? Repeat work means the client sees you as a resource they call on. One-off work means they're shopping. The market is consolidating around embedded, advisory relationships. The one-shot consulting firm is a project vendor now.
Bottleneck 5: Economics of Replication. What would it cost the client to replicate this work in-house versus paying you? If it's cheaper to hire a contractor for $80K and run it once a year, your pricing is wrong or your work is commoditized. If the client would need to staff someone full-time to match your output, you own scarcity.
Score each service line. Services hitting 4 or 5 bottlenecks are defensible. Services hitting 2 or fewer are at risk.
What's Changing Inside the Big Firms
McKinsey deployed Lilli:a generative AI platform:firmwide in July 2023. Within a year, 72% of their team was using it daily.[^mckinsey] The work didn't disappear. The staffing model changed. They compress what used to take a team of four into what a senior plus an AI agent can handle. That cost savings doesn't get passed to the client. It gets eaten as margin, or it gets invested in higher-end advisory.
The use they're accessing is a classic manufacturing insight: reduce the number of low-value touches, increase the strategic guidance. That is the business model of the surviving consulting firm. What it means on your pricing sheet is this: If you're paid by the seat-week or the analyst-week, your revenue goes down. If you're paid by the business outcome, it goes up. The client doesn't care how many people you used. They care whether the answer was right.
The firms that used to compete on "we have the smart people" now compete on "we have the smart people plus the apparatus." The apparatus is the AI-driven intake, the templated delivery pipeline, the proprietary data feeds, the system integration. That apparatus is cheaper to build than you think. It's more valuable than you think.
How to Build the Moat Back
1. Own the vertical. Stop being a general strategist. Be the strategist for insurance-tech companies, or orthopaedic supply chains, or regulatory compliance in biotech. Build a practice that is synonymous with that vertical. That means 10+ years in, deep relationships, knowledge of the failure modes the industry hasn't yet named.
2. Commercialize your IP. Take your methodology. Give it a name. Document it. Build a practice around it. Make it the thing people hire you to run. Not "strategy consulting." "The [Your Name] Framework for post-M&A integration in healthcare." Proprietary methodologies sell at multiples. Generic advisory doesn't.
3. Build a data moat. Run your own surveys. Build your own benchmarks. Develop proprietary indices. Keep that data private:it's use, not a free lead-gen tool. Your largest clients should pay for access to your data. Your data should be irreplaceable enough that they renew.
4. Become a system. Don't deliver reports. Deliver ongoing monitoring, dashboard access, quarterly reviews, real-time alerts. Move the client from "we hired you for this project" to "we pay you each quarter to watch our engine room." That economics is completely different. Margin is higher. Retention is higher. The client can't leave without losing visibility.
5. Own the implementation. The moment you hand off your analysis, your use ends. If you stay through implementation:even as an advisor, even at a lower fee rate:you own the truth of what happened. The client can't point to your analysis and say "it didn't work." They know it did, because they watched you guide it. That's a moat.
6. Invest in delivery compression. Build your own agent-assisted pipeline. Wire intake into drafting into review. The math is simple: if a team of two with a solid AI stack can produce what used to take four people, your margin and your margin per dollar of revenue go up. That is not automation replacing humans. That is use. You keep the partner margin. You reduce the staff burden. You price the same. You pocket the difference.
The Verdict
Claude and GPT didn't kill consulting. They killed the consulting firm that competed on slide-deck quality, generic frameworks, and fast turnaround. They killed the commodity end of the market.
They also opened the door for small firms to compete on moat-bearing work at institutional prices, because the overhead of delivery just dropped. If you own vertical depth, IP, proprietary data, or system status, you can now charge like a specialist and actually deliver like one. The cost of delivery compression made that math real.
If you don't own those things, the answer is simpler: build them, or exit into advisory roles at bigger firms where the pipeline, the brand, and the data density make up for thin methodology.
The firms that pass the "Can your client build it?" test are the ones that answer "no" confidently. They have something the client can't replicate:not talent, not process, not framework, not hours billed. Something real. That's the only consulting left that matters.
FAQ
Q: How fast will my current clients start asking "can't we just use AI?"
A: The operations person at your client has already asked. Your champion inside that company has heard the question and brushed it off. The question is not "will this happen?" but "when does it move from the operations office to the CFO's office?" Bet on this: Two years from now, every procurement team under $100M in revenue will have at least one person whose job is "figure out what we can pull in-house with AI instead of buying it from consultants." That person is already working. Get ahead of it now.
Q: Can I use this audit on my existing clients?
A: Yes. Run it on your top 10 clients by revenue. For each, map which services scored high (4-5 bottlenecks) and which scored low (1-2). The low-scoring services are vulnerable. Either rebuild them into something defensible, price them as commodities and accept margin compression, or phase them out. The high-scoring services are what you reinvest into. That's your moat. Double down on it.
Q: My whole firm is built on fast-turnaround strategy reports. Am I already dead?
A: Not dead. Disrupted. You have a three-year window before the market fully reprices your work. Use it to build vertical expertise, develop proprietary methodology, or acquire data assets. The firms that will not make the transition are the ones that treat this as a pricing problem. It's not. It's a value problem. Solve the value problem first.
Q: What does "system of record" actually mean?
A: It means you become infrastructure. Quarterly business reviews. Ongoing market monitoring. Real-time alerts on competitive moves. The client doesn't bring you in for a project:they bring you in monthly or quarterly because they trust you to watch the engine room while they handle the deck. You're not selling hours. You're selling certainty. The fee structure is retainer-based, and it renews because not knowing what you're watching is scarier than paying the fee.
Q: Should I build the AI stack myself or hire an agency?
A: Build it yourself, or hire someone to build it with you:not for you. The stack is your delivery methodology now. It's part of your IP. If you hand it to an agency and they build it, you don't own it, understand it, or control how it evolves. You'll be asking them for changes instead of shipping them yourself. Own the tool chain. Competence beats credentials.
[^hbr]: Harvard Business Review, "AI Is Changing the Structure of Consulting Firms," September 2025, https://hbr.org/2025/09/ai-is-changing-the-structure-of-consulting-firms
[^valuation]: Iconic, "Valuation Multiples for Consulting Firms: What Drives Price in 2026," July 2026, https://iconic.co/blog/valuation-multiples-for-consulting-firms/
[^mckinsey]: McKinsey & Company, "Rewiring the way McKinsey works with Lilli, our generative AI platform," https://www.mckinsey.com/capabilities/mckinsey-digital/how-we-help-clients/rewiring-the-way-mckinsey-works-with-lilli
[^used]: The used Years, "McKinsey Has 25,000 AI Agents. Build Your Own Stack," June 2026, https://www.theusedyears.com/ai-workflows/boutique-ai-agent-delivery-stack-consultants
[^tly2]: The used Years, "McKinsey Has 25,000 AI Agents. Build Your Own Stack," June 2026 - on AI agent delivery workflows and boutique stack architecture
Jeff Barnes is the founder of DEMG.ai and Digital Evolution Marketing Group. He has no personal position in any company, fund, or platform named in this article. DEMG.ai provides marketing systems and education for owner-operators, not investment advice. Past performance does not guarantee future results.
