Why The Industry Is Stuck

Chapter Eight told you what is wrong. This chapter tells you why.

The camp dynamic that Chapter Eight critiqued — the AI camp and the old-school SEO camp creating false choices for contractors — is the visible surface of a deeper problem. The contractor reader who has been frustrated with their current marketing partner deserves to know what is actually happening inside that partner’s business, and what is actually happening inside the broader industry, that has produced the situation Chapter Eight described. The problem is not just one of bad advice. The problem is that real cognitive, structural, and economic constraints are keeping operators stuck in the previous decade’s playbook, and most of those constraints are not the result of cowardice or bad intent. They are the result of how marketing agencies were built, who has historically led them, and what the actual economics of running one of those agencies look like in 2026.

I want to be honest about this, because the contractor reader who is going to make a good decision about their next advisor needs to understand the constraints their current advisor is operating inside. Empathy without endorsement. The constraints are real. The path through them is also real. Some operators have made it through. Most have not. The four problem states this chapter describes are the reasons most have not, and the contractor reader who recognizes their current advisor in any of these states should not necessarily fire them — the contractor reader should understand that their current advisor is operating inside a hard situation, and should make an honest decision about whether the situation is one the contractor can afford to keep paying for.

Compounding Rates Of Change

The first problem is cognitive.

The human brain is not built to project compounding curves accurately. We can intuit linear change well — a salesperson who closes one more deal each quarter than the previous quarter is doing observably better, and we can extrapolate that improvement forward without effort. We cannot intuit exponential change. The technology that doubled in capability every year for the previous five years feels, to most observers, like it has been changing at a roughly steady pace, because each year’s change felt comparable to the previous year’s change. The cumulative result of those compounding doublings is something that most operators only register after the curve has gone vertical and the gap between where they are and where the leading edge is has become impossible to close. By the time the operator can see the gap clearly, the gap is already too large to bridge with the resources they have available.

That cognitive limitation is not a character flaw. It is how the human brain works. Every operator reading this book has it. I have it. The reason I have been less surprised by the AI transition than most of my peers is not that I have better cognitive equipment. The reason is that I have been spending three years deliberately watching the curves and forcing myself to project them forward, against my own intuition, on the assumption that what I expected to happen was probably going to be slower than what would actually happen. That deliberate exercise of distrust-your-intuition-on-exponentials is what has kept me in the right rough position. The operators who are stuck in the previous decade’s playbook are not stuck because they are dumb. They are stuck because they trusted their intuition about a kind of change their intuition was never designed to evaluate.

This problem compounds with the rest of the problems this chapter is going to describe. The operator who cannot project the rate of change cannot evaluate whether the investment required to make the AI transition is worth making. The operator who cannot evaluate the investment cannot make a confident decision. The operator who cannot make a confident decision defers, hedges, or stays the course. Compounding rates of change is therefore the cognitive root of the camp dynamic Chapter Eight critiqued. The camps exist because operators are trying to navigate an environment they cannot accurately project, and the camps offer the false comfort of a confident position when the honest position is uncertainty. That is why the *AI is here, get in or get out* camp and the *AI is hype, stay the course* camp both feel emotionally satisfying to their participants. Both camps offer certainty. The actual situation does not have certainty in it.

I want the contractor reader to register this clearly. If you are talking to an advisor who is offering certainty about what AI is going to do, you are talking to an advisor who is wrong. The honest advisor in 2026 is the advisor who tells you the directions are clear and the specifics are uncertain. Anyone who is selling you certainty is selling you a posture, not a forecast.

Constraints And Investment

The second problem is structural and economic.

Most marketing agencies operate on tight margins. The agency CEO who is contemplating the AI transition is not making a decision in a vacuum. They are making a decision against a backdrop of payroll obligations, client deliverable commitments, lease payments, and debt service — the unglamorous economic reality of running a services business at scale. The investment required to become AI-enabled or AI-first, done correctly, is enormous. It is more than money. It is skill, which the agency may not have on staff and may not be able to hire affordably. It is risk, because the AI infrastructure being built today may be obsolete in eighteen months and the investment may not return before the technology shifts again. It is people, because the existing team that has been running the legacy business may not be the team that can run the post-transition business, and the operator may have to make hard decisions about people who have been loyal for years. It is brand, because every misstep during the transition damages the brand the operator has spent years building, and the early movers are sometimes punished for being early when their first attempts ship before they are ready.

None of these are easy decisions. The agency CEO who is staying the course is sometimes making the rational decision given their constraints. They cannot afford to make the AI investment without putting the legacy business at risk. They cannot afford to lose key team members during the transition. They cannot afford the brand damage that comes with shipping half-built technology. Their constraints are real, and the constraints are why they are stuck. The contractor reader who is frustrated with their current advisor for not having made the AI transition needs to understand that the advisor may not be wrong. The advisor may be operating inside constraints that genuinely do not permit the investment, given the resources the advisor has available.

Our clients can see this structure from the outside, without any of the vocabulary in this chapter. John Dean — whose full account is in Voices from the Clients — watched the agencies attacking our direction and worked out the economics on his own:

The resistance I was hearing from other agencies wasn't necessarily because they genuinely believed Jennifer was wrong. Many simply didn't want to make the enormous investment — in time, money, education, and technology — that would be required to reinvent their businesses for an AI-driven future.

That understanding does not change the contractor’s decision. The contractor still needs an advisor who has made the AI investment, regardless of why other advisors have not. The contractor’s business is not improved by the contractor’s sympathy for an advisor whose constraints prevented them from delivering what the contractor now needs. But the understanding does change the moral register of the decision. The contractor who is firing their current advisor because the advisor cannot deliver the work the contractor now needs is not making a moral judgment about the advisor. The contractor is making an operational decision about what the contractor’s business requires. Those are different decisions, and treating them as different decisions is the difference between a professional relationship that ends with respect and one that ends with bitterness on both sides.

There is also a category of agency leader the contractor reader should know about specifically. Most marketing agencies are not owned by people with developer backgrounds. Traditional marketers are, on average, more creative than technical. The agency CEO who can lead a creative team is not necessarily the agency CEO who can lead the technical rebuild that AI deployment requires. That is a structural fact about who has historically founded marketing agencies, and it is one of the reasons the industry is genuinely struggling with this transition. The creative-not-technical leadership that built the agencies of the previous decade is not the leadership that can rebuild them for the next decade, and the operator who recognizes this gap in their own background and hires for it is operating well beyond what most of their peers are doing. Most are not.

Superficial Knowledge

The third problem is more dangerous than admitted ignorance.

There is a category of operator who has used ChatGPT, Claude, Cursor, or Lovable, and who has therefore concluded that they understand AI. They do not. They understand the consumer-facing surface of AI tools, which is a tiny fraction of what AI infrastructure actually consists of. The operator who thinks AI is ChatGPT does not understand vector databases. They do not understand RAG — retrieval-augmented generation — or CAG — cache-augmented generation. They do not understand data pipelines, terminals, hardening, security. They do not understand the difference between a single-page application and a server-side-rendered website, and they do not understand why that difference matters for whether AI agents and bots can index a contractor’s digital footprint correctly. They have superficial understanding mistaken for fluency, and their superficial understanding is more dangerous than admitted ignorance because they are confidently making decisions on a knowledge base that does not exist.

The contractor reader who is talking to an advisor who claims AI fluency should ask specific technical questions. What is your retrieval architecture for client knowledge? How do you handle context-window management for client-specific agentic workflows? What is your vector-database posture, and what model are you using for embeddings? How do you secure agentic systems against prompt injection? What is your client data isolation architecture? The advisor who can answer these questions is operating from a real knowledge base. The advisor who cannot answer them, or who deflects with confidence-projecting language about *AI capabilities* and *AI-powered solutions,* is operating from the superficial layer that does not actually qualify them to do the work. That is a hard test. Most advisors will fail it. The contractor reader who is going to make a good decision about their next partner needs to be willing to ask the hard questions and willing to recognize the difference between a substantive answer and a confidence-projecting non-answer.

I want to be specific about why this matters operationally. The contractor whose website is being rebuilt by an advisor who does not understand the SPA-versus-SSR distinction is going to end up with a beautiful-looking website that AI agents cannot index. The contractor whose data pipelines are being built by an advisor who does not understand security hardening is going to have their customer data exposed in the next breach cycle. The contractor whose agentic workflows are being deployed by an advisor who does not understand prompt injection is going to have a customer-service agent that can be hijacked into providing free quotes or revealing internal pricing. These are not theoretical risks. These are the operational realities of working with an advisor whose AI fluency is superficial. The contractor pays the cost. The advisor moves on to the next client.

The honest path is to find an advisor who has the technical depth to actually do the work. They exist. They are also rare, which is part of why most contractor businesses are working with advisors who do not have the depth. The supply of genuinely qualified AI-fluent agency operators is much smaller than the demand. The contractor who finds one and keeps them should hold onto the relationship like the asset it is.

Rookies With The Wrong Tools

The fourth problem is the new entrants who do not know what they do not know.

There is a wave of new operators entering the trades-industry technology space who have learned to use modern AI-assisted development tools — Vercel, Lovable, Cursor, the various low-code-and-AI-augmented website builders — by watching TikTok and YouTube videos. I want to be careful with this section because new entrants are not the problem in principle. New entrants are how the industry refreshes its talent base, and many of the operators who started watching tutorial videos in 2023 will be the senior operators of the 2030s after they have done the deeper learning the work requires. The problem is not new entrants. The problem is new entrants shipping production work for paying clients before they have done the deeper learning. The rookie who is studying React, learning indexing architecture, building toward technical depth alongside their tool fluency is going to be a qualified operator in a few years. The rookie who is selling a contractor a Vercel-built website this quarter is shipping a product the contractor will pay to clean up later. The distinction matters.

The rookies shipping production work today are confident. They are productive. They can ship a website in a few hours that looks impressive in screenshots and demo videos. They have no idea that the React JavaScript website they just built is a client-side-rendered single-page application, and they have no idea that this means Google’s indexing systems and the AI agents that are increasingly evaluating contractor websites cannot read the content properly. They do not know what they do not know. The website looks amazing. The technical foundation underneath it is unsuitable for the use case the contractor needs it to serve.

The contractor business owner who hires the rookie or buys their *amazing* app is going to take real damage when the limitations surface. The damage will surface when leads stop coming in because the website is not being indexed. The damage will surface when the security holes in the rookie-built application get exploited. The damage will surface when the AI agents evaluating the contractor’s digital footprint return inadequate or wrong information about the business because the website was not built to be machine-readable. The damage will surface when the contractor’s competitor, working with a more experienced operator, starts winning the AI-mediated conversations the contractor cannot participate in.

There are BS apps being launched every day that look amazing in their marketing material. They were built on a weekend. They are insecure. They have no architecture for scale. They have no security hardening. They have no integration with the systems contractor businesses actually depend on. The contractor who buys one is paying for a product that will fail in production, and the contractor will not understand why. The rookie who built it will move on to the next thing. The contractor will be left with a broken stack and no recourse.

I want to register this clearly because the trades industry is going to have to learn the hard way real fast. The lessons are coming. The contractor businesses that get burned by working with rookies in 2026 are going to have to clean up the damage in 2027 and 2028, and the cleanup will be more expensive than doing it right would have been to begin with. The contractor reader who is being pitched by a confident young operator with a portfolio of beautiful-looking websites built in Vercel should ask the hard technical questions — the same questions from the previous section. SPA versus SSR. Indexability. Security hardening. Data architecture. Long-term maintenance. If the answers are confident but technically thin, the contractor is talking to a rookie. The contractor does not need a rookie.

The Selection Problem

The previous four sections described who is stuck. This section describes one of the hardest tasks the operator who is not stuck still has to perform.

The contractor or operator navigating the AI transition has to make selection decisions — who to hire, who to trust, who to learn from, who to follow, who to ignore. Each decision is consequential. Each one carries downstream risk. The operator who selects the wrong vendor ends up with a broken stack and wasted capital. The operator who selects the wrong mentor ends up absorbing bad judgment that compounds across years of subsequent decisions. The selection problem is not the same kind of problem as the four states this chapter has already described — the previous four are conditions an operator might be in. Selection is a task an operator is performing. But selection is hard enough, and getting it wrong is expensive enough, that the chapter would be incomplete without addressing it directly.

Vendor selection and mentor selection are related but they are not the same. Vendor selection is transactional and replaceable. The contractor who hires the wrong agency can fire the wrong agency in twelve months and try again. The cost is real — wasted capital, lost time, brand damage from poorly-executed work — but the relationship ends and the contractor moves on. Mentor selection is durable and shapes how the operator thinks. The contractor who absorbs bad judgment from a mentor over five years has internalized a set of operating assumptions that are now hard to dislodge, even after the contractor recognizes the mentor was wrong. The damage from a bad mentor is harder to identify and harder to undo than the damage from a bad vendor. Operators tend to be more careful about vendors than mentors, which is exactly backwards.

On vendor selection, the framework I am about to give you in the chapter close is the tool I want you to use. Four questions, four diagnostic frames. Run any prospective vendor through that framework before signing a contract. Run your current vendor through it on a regular review cadence. The vendor who passes the framework today may not pass it in two years if the technology has moved past their capabilities. Re-running the assessment is part of the discipline. Long-term vendor contracts in the AI era are particularly dangerous because the technology shifts faster than the contracts allow operators to renegotiate. The contractor reader should be wary of any vendor who is asking for multi-year commitments without renegotiation clauses tied to specific capability milestones. The vendor who needs the long contract is the vendor who is not confident they can keep delivering.

On mentor selection, the criteria are different and most operators get them wrong. The mentor most operators select is the visible mentor — the operator with the conference keynote, the podcast presence, the social media following, the published book. Visibility is a signal of marketing investment. It is not a signal of operational competence. The mentor whose business has actually navigated the kind of transition the contractor is trying to navigate is the mentor whose advice is worth paying for. The mentor whose business is built on selling advice rather than on operating in the category they are advising about is, frankly, more likely to be the operator profiting from confusion than the operator resolving it. The contractor reader should be especially careful of any mentor or coaching program whose business model depends on the contractor staying in a state of needing more advice. That is a structural conflict of interest. The mentor whose financial incentives are aligned with the contractor actually solving the problems they are being mentored on is rare. The mentor whose financial incentives depend on the contractor coming back for more is common.

There is a specific pattern I want to name because it is the most common failure mode I have observed in the trades industry. Contractor A asks Contractor B for a referral to a mentor or vendor. Contractor B refers Contractor A to the mentor or vendor Contractor B is using. Contractor A signs up. The problem is that Contractor B is often not qualified to evaluate whether the mentor or vendor is actually delivering value, because Contractor B has the same compounding-rates blindness, the same superficial-knowledge limitations, and the same constraint-based decision-making as the rest of the industry. The referral is therefore not a quality signal. It is a network signal. Contractor B uses this mentor, Contractor A trusts Contractor B, therefore Contractor A uses this mentor. The mentor’s actual quality has not been evaluated by anyone in the chain. That is how mediocre mentors and mediocre vendors get embedded in the trades industry over years — not through any individual operator deliberately choosing badly, but through the referral network propagating the same selections regardless of underlying quality.

The contractor reader who is making selection decisions in 2026 should default to skepticism about referral-based recommendations and instead apply the evaluation framework directly. The contractor who runs the framework against a referred vendor and finds the vendor failing the technical-depth question or the investment-honesty question should not assume the referring contractor knows something they do not. The referring contractor probably does not. The selection that propagates through the referral network is not necessarily the right selection. Apply the framework. Trust the framework over the referral.

Four problem states.

Compounding rates of change that the human brain is not built to project. Constraints and investment realities that keep capable operators stuck in postures that are no longer adequate. Superficial knowledge mistaken for fluency by operators who confidently make decisions on a knowledge base that does not exist. Rookies with the wrong tools who can ship beautiful surface but cannot deliver the foundation underneath. These four problem states overlap, reinforce each other, and produce most of the dysfunction the contractor reader will encounter when looking for an advisor or partner during the AI transition.

The trades is going to have to learn the hard way real fast.

Before this chapter closes, I want to give you something concrete you can use — an evaluation framework for assessing the advisors and partners in your own market who may not be named in the chapter that follows but who you will need to make decisions about regardless. The framework is a set of questions to ask, paired with how to interpret the answers.

The trades is going to have to learn the hard way real fast.

Ask your current advisor or any prospective advisor: how do you think about the rate of AI evolution, and how does that thinking change what you build for clients? The answer that signals competence is one that acknowledges the projection problem honestly — the advisor who tells you they are watching specific signals, building for short time horizons, and updating their assumptions quarterly is operating from a real cognitive discipline. The answer that signals one of the four problem states is confident certainty about what AI is going to do or dismissive certainty that AI is overhyped. Either certainty posture is a signal of compounding-rates blindness.

Ask: what is your investment posture toward AI infrastructure, and what tradeoffs are you currently making between supporting your legacy business and building for the AI transition? The answer that signals competence is one that owns the difficulty honestly — the advisor who tells you they are running both systems concurrently, that the investment is hard, that the tradeoffs are real, is operating with operational honesty. The answer that signals constraints-and-investment trouble is one that pretends the transition is easy or that the advisor has fully completed it. The advisor who claims to have completed the transition has either not actually done it or does not understand how much of it remains.

Ask: what is your technical architecture for client AI work? Specifically, what is your retrieval-and-context architecture, what are your security practices for agentic systems, and what is your approach to indexability for the websites you build? The answer that signals competence is one that uses specific technical vocabulary correctly — the advisor who can explain RAG versus CAG, who can articulate why client-side rendering matters for AI indexability, who can describe their security posture against prompt injection, is operating with real technical depth. The answer that signals superficial knowledge is one that uses the vocabulary loosely, defers to *AI-powered* generalities, or projects confidence without specifics. If the advisor cannot answer the technical questions in technical language, the advisor does not have the technical depth the work requires.

Ask: what is your tooling philosophy and how do you make build-versus-buy decisions for client deliverables? The answer that signals competence is one that distinguishes between tools good for prototyping and tools appropriate for production, between fast-shipping demos and architecturally durable products, between trendy frameworks and frameworks that will still be supported in three years. The answer that signals rookies-with-wrong-tools is enthusiasm for whichever AI-assisted development tool the advisor learned most recently, paired with confidence that the tool can solve any client problem. The advisor who has not done the work of distinguishing prototype-grade tooling from production-grade tooling is going to ship the contractor a prototype and call it production.

Four questions. Four diagnostic frames. The contractor who runs through this checklist with their current advisor or with a prospective advisor will know within ninety minutes whether the advisor is operating inside one of the four problem states or whether the advisor has navigated past them. The framework is not perfect. There are advisors who will give competent-sounding answers without having the underlying competence to back them up, and there are advisors who will struggle with the language without lacking the underlying capability. But on average, in the contractor reader’s actual experience, the framework will produce more accurate signal than the contractor is currently getting from any other source. Use it.

There is one more thing I want to give you before this chapter closes, and it is something the previous panel of marketing experts I assembled mentioned was missing from my treatment elsewhere in this book. The four problem states this chapter has named are real and serious, and they are also partially solvable through community-based learning. The contractor who is part of a community of operators who are openly sharing what they are learning, what is working, what is not working, and what they are watching is partially insulated against compounding-rates blindness because the community is doing the projection collectively. The contractor who is part of a community can stress-test technical claims against operators with deeper expertise than any individual contractor will have on their own. The contractor who is part of a community can pool resources and reduce the per-operator cost of the AI transition by sharing tools, sharing learnings, and sharing the cost of the experimentation that the transition requires.

I run one of these communities. JustStartAI is a community of more than a thousand contractor members who are sharing what they are learning about AI in the trades, building together, and contributing what they figure out for the broader benefit of operators who are trying to navigate this transition. I am naming JustStartAI here not as a sales pitch — the community has been free for its members and is intended to be — but as a proof point that the community-based approach the engineering community has been using for thirty years can work in the trades industry too. The contractor reader who is not currently part of any community of AI-learning operators is operating without one of the most important defenses against the four problem states this chapter has described. Find a community. JustStartAI is one option. There are others. The specific community matters less than being inside one. Operators who are learning together navigate this kind of transition substantially better than operators who are learning alone.

If you are going to engage with JustStartAI specifically, the entry point is intentionally simple. The community lives at JustStartAI.io. New members fill out a short application that asks who they are, what kind of trades operation they run, and what they are currently trying to figure out about AI. The application is not a screening filter — we accept almost everyone who applies — it is a way to make sure new members get connected with the existing members who are working on similar problems. After acceptance, members get access to the community channels, the regular live calls, the prompt library, the case studies that other members have shared, and the working groups that form around specific problems contractors are trying to solve. The work is done by the members, not by me. I am one voice in a community of operators who are figuring this out together. That is the version of community-based learning that I am asking the contractor reader to consider participating in. Whether you join JustStartAI or whether you find a different community that fits your specific operational profile better, the point is the same. Do not try to figure out the AI transition alone. The operators who are succeeding at it are operators who are inside a community of other operators doing the same work.

The agency that has built operational discipline for living inside the four problem states without being captured by them is the agency the contractor reader should be looking for. CI Web Group is one such agency. There are others. None of us have figured the AI transition out. None of us are claiming to have. What the agencies in this category have built is the discipline to ship fast, gather data, and pivot when the data tells us to — instead of staying stuck in the postures the previous four sections described. The next chapter names five operators — some of them agencies, some of them platform builders, some of them industry voices — who have built that discipline with operational seriousness. The chapter that follows is the *here is who built the discipline* chapter that this chapter’s diagnostic prepared you to read.

Now, the coalition.