AI Sameness & The Firms That Sell What They Know

Put four proposals from four competing firms side by side, each written this year with help from AI.  Each one is better than what each firm would have sent three years ago, with a cleaner structure, a sharper summary, and fewer errors. Read together, though, they are hard to tell apart: each frames the client's problem the same way, proposes the same three-phase approach, and closes with the same reassurance.

That is AI sameness.

When competitors use the same tools in the same default ways, their work drifts toward the same typical answer: the same voice, the same structure, the same ideas, and eventually the same blind spots. For most companies, that is a nuisance. For firms whose product is knowledge, it is a business problem. Accounting, legal, and consulting practices; engineering and architecture firms; software and technology companies; and marketing and creative agencies all sell some version of the same thing: what they know or see that the other firm doesn't.

We recently took this topic to a roundtable of accounting firm leaders, and what we told them applies to any firm that sells what it knows. Your leadership approved the AI tools that are right for you. What most firms haven't decided is how those tools will change the firm: which work gets standardized, which skills stay in people's hands, and whose voice the work carries. If nobody decides, the defaults built into the software decide for you.

Individually Better, Collectively Alike

The research backs this up. In a 2024 study in Science Advances, Anil Doshi and Oliver Hauser gave some writers AI-generated ideas and gave others none. Judges rated the AI-assisted stories as more novel and more useful. Those same stories were also measurably more similar to each other. A field experiment with management consultants, run by researchers at Harvard Business School, found the same pattern on realistic consulting work: the quality of ideas went up, and their variety dropped markedly. One of its co-authors, Ethan Mollick, author of Co-Intelligence, is among the most optimistic voices on AI at work.

The improvement hides the sameness. Inside your firm, every signal says AI is working. Work is faster and cleaner, and newer staff turn in drafts that read like a senior's. You only see the sameness from the client's side of the desk, where your work sits next to a competitor's. (NOTE: Just compare a week's worth of sales emails in your inbox and you'll get the point).

A language model answers with a likely response based on what it learned. In practice, that means a typical answer, written well. Five things push competing firms toward the same answer:

1. The leading models learned from heavily overlapping material.
2. They were tuned on human ratings to make them helpful and polished, and research shows that tuning also narrows the range of what they produce.
3. Most business software, whether for proposals, research, design, code, or marketing, runs on a handful of those models with similar templates on top.
4. Most firms we see set up AI with its defaults, without their own brand standards or firm-wide context, which means the branding and context come from the tools.
5. People accept the first draft.

The first three come with the tools. The last two are decisions. Competitors who think they bought different products end up depending on the same few engines, and their improvements all point in the same direction.

Professional firms borrowed each other's templates long before AI. What is new is how fast the work converges, and the fact that everyone is now drawing from the same source. None of this means the tools can't produce distinctive work. Give them your firm's method, evidence, and point of view, and they will produce something much closer to your firm. Sameness is what happens when nobody decides to supply those things. It is also different from what researchers call model collapse, where models degrade after training on AI-generated content. Sameness needs no degradation at all. It happens with the tools working exactly as designed.

Why Firms That Sell Knowledge Feel It First

Every company that adopts AI is exposed to some version of this. Firms that sell knowledge are exposed differently, because what converges is the product itself. A manufacturer whose marketing copy sounds like its competitors' still ships a different part. A firm whose advice, design, analysis, or creative work looks like its competitors' has lost the thing clients were paying for. Here is how it shows up in three kinds of firms.

Accounting, Legal, & Consulting

The deliverable is the judgment, so sameness shows up in proposals, memos, and recommendations. Not all of that is a problem. Audit and tax compliance are governed by standards, so the work is consistent from firm to firm, and nobody wants a creative audit opinion. The exposure is in the advisory work clients pick you for, and in the apprenticeship that builds your next generation of advisors.

Software & Technology Services

Sameness shows up in the product and in how it is sold. AI coding tools tend toward similar interfaces, structures, and wording. Because our practice helps clients select technology, we get a steady flow of vendor pitches, and more of them read alike every month: the same opening compliment, the same bolded labels, the same emoji used as bullets, the same gradient title slide and grid of icons. None of those choices is wrong on its own. What gives them away is that they show up together, from companies that each think they are standing out.

Marketing, Media, & Creative Agencies

Here the work is the voice, so the exposure is the most direct. The Science Advances study measured exactly this kind of work. Agencies also carry a second exposure. When an agency's tools pull toward the typical answer, its clients' brands get pulled along with it.

That second exposure applies to every firm on this list, and to their clients. Our team recently tested a client's application before it went live. It was a site the client's own vendors use, built by an outside developer largely by prompting an AI tool and accepting what came back. The layout, the wording, and the color scheme all carried that tool's defaults. To its vendors, our client now looked like every other application built the same way. The client never made that choice. Its suppliers' defaults made it. Your sameness can come in through your suppliers just as easily as through your own staff.

What It Costs The Firm

The cost travels two paths, and both end at what the firm is worth. The first runs through the market. When clients can't see a difference between firms, they compare them on price, and lower prices come out of margin. A firm whose earnings rest on work anyone could produce is also worth less when it brings in a successor or outside capital, because a buyer pays for what is hard to replicate. That chain is reasoning. No one has yet measured how much sameness lowers fees, and we won't put a number on it until someone does.

The second path runs inside the firm, and it is the one we would watch most closely. Four things tend to happen while the work is getting better.

Capability Atrophy

Skills fade when the tool does the work, and newer staff never build them. Lisanne Bainbridge found the same thing in automated industrial control rooms in 1983. In a knowledge firm, those skills are where judgment comes from. Who in your firm could rebuild this work if the tool went away?

Methodology Overwrite

The tool's template quietly becomes your method, and your competitors bought the same template. Is your method still yours?

Verification Erosion

You can't check what you can no longer do. Research on automation bias finds that people stop checking carefully when a system is usually right, experts included. Research on shared models adds that decision-makers relying on the same one tend to get the same cases wrong. Who checks your AI-assisted work, and could they still do that work themselves?

Dependency Accretion

Switching costs pile up as tools get built into workflows and training, and nobody approves that growth. If your main AI supplier changed its price or its product next quarter, what would you do?

Put together, these take away the firm's ability to be different, to notice it has stopped being different, and to change course. In resilience terms, the firm loses its recovery capacity. When clients start rewarding distinctive work again, it won't have the skills, methods, or people to deliver it.

Why Your AI Governance May Not Catch It

Most firms adopting AI have done the responsible things: an acceptable-use policy, a security review, a vendor assessment, maybe a governance committee. In some of the risk committees we facilitate, AI oversight centers on the technology: what it costs, which tools to use, and whether they are secure. Those are the right questions, but they are incomplete. Risk committees are watching the AI, not the company.

The established AI governance frameworks do cover risk to the organization, mostly in the form of the AI going wrong: errors, bias, security, privacy, reputation, and compliance. They say much less about what happens when the AI works exactly as designed, and the firm slowly starts to look like its competitors, loses skills it used to have, and builds up dependency it never chose.

None of this is new to risk management. Concentration on a few shared suppliers, a control that quietly stopped working, a capability you can no longer recover: boards and risk committees already know how to govern each of these. Few have pointed that discipline at what AI adoption is doing to their own firm.

Make Six Decisions, & Name An Owner For Each

You should be using AI, and you probably already are. The next step is to take back the decisions the defaults are making for you. There are  six, and each one needs a named owner.

None of these needs a technology budget. They need leadership time. Plenty of firms will choose the default for most of their work and protect a few things on purpose, and that is a perfectly good decision. The risk is ending up with the default without ever deciding.

Start With Three Questions

Three questions will tell your leadership team where it stands:

1. Could a client tell our last proposal, design, or campaign from a competitor's with the logo removed?
2. Which of the six decisions has a named owner today?
3. Is our risk committee asking what AI is doing to the firm, as well as what it costs and whether it is secure?

The firms that stay unmistakable will use AI as heavily as anyone. They will also go back and decide the rest on purpose. Pick one of the six decisions this week and decide who owns it.

About GBQ Business Technology Solutions

GBQ's Business Technology Solutions practice empowers the growth of our clients across six disciplines: risk management, cybersecurity, IT governance, AI and automation, data and analytics, and business systems.

If this article raised questions that you cannot yet answer about your own strategy, the first place to start is with a conversation. GBQ’s Business Technology Solutions team can assess where your organization stands today, prioritize use cases worth your investment, and help build the governance that lets you prove what AI is returning.

Net Effect is a biweekly column written by members of GBQ's Business Technology Solutions team and published in the firm's Bottomline newsletter. Let us know if you have a technology question, and we can address it in a future column.