October 30, 2026

AI for the Over 40 – Week 37: Do We Still Need Consultants?

Years ago, we made a decision that most of our competitors thought was crazy. We started blogging heavily. Not thought leadership pieces or gated whitepapers that required an email address, but actual technical content: how to configure posting groups, how to troubleshoot common errors, how to set up dimensions. This was the kind of content that clients were calling our support desk to get and paying us by the hour to receive.

Our peers told us we were giving away the store. If clients could find the answer themselves by reading a blog post, why would they call us? We were cannibalizing our own support revenue.

My contention was simple: any answer a client could find by reading a blog post was not a business model worth preserving. If the only reason they were calling us was that the answer was hard to find, not that it was hard to understand or implement, then we were charging for access rather than expertise. I wanted us to give away the easy answers for free and focus on delivering the hard stuff that justified what we charged.

It worked. The blog drove more business than the support calls it replaced. Clients who found answers themselves came to trust us more, not less. And when they hit problems they could not solve with a blog post, the hard stuff, they called us first, because we had already proven we knew what we were talking about.

I am telling you this story because AI just did to consulting what our blog did to support desk revenue. Except it did it at a scale I could not have imagined, and it did not stop at the easy answers.

Why AI changes the value of consulting

When we started blogging, the line between “easy answers we should give away” and “hard work worth paying for” felt clear. Clients could look up how to configure a posting group. They still needed us to diagnose why their financial statements were not reconciling across five entities with intercompany transactions in three currencies. The knowledge was accessible. The judgment was not.

AI has moved that line dramatically.

A client can now ask AI to explain posting groups, walk them through configuration step by step, troubleshoot errors in real time, and answer follow-up questions with more patience and availability than any consultant. That is the easy answer layer, the work our blog was already giving away. No surprise there.

But AI also helps clients think through multi-entity consolidation strategies. It can analyze their chart of accounts and suggest restructuring. It can explain the implications of different dimension architectures. It can generate implementation plans with task dependencies, resource estimates, and risk assessments. It can review their business processes and identify inefficiencies. And it does all of this instantly, at any hour, at effectively zero marginal cost.

That is not the easy stuff anymore. That is, work consultants get paid serious money to deliver.

Last week, I argued that AI does not replace developers; it concentrates and changes what they do. The steelman against developers was strong, but it hit concrete walls: single-user limits, prototype-to-production gaps, and budget reallocation from MVP toward the real vision.

The case against consultants is harder to dismiss. The walls are less visible, and the value proposition is more intangible. I need to be honest here: the consulting model has vulnerabilities that go beyond what AI can do. Some of those vulnerabilities are things AI exposes. Others were always there.

The case against consultants: what AI replaces

What clients can now do without a consultant

The traditional consulting model is built on a knowledge asymmetry. The consultant knows more than the client about the product, about best practices, about what works at other companies, about the workarounds, configurations, and design patterns that turn generic software into a system that fits a specific business. The client pays for access to that accumulated knowledge.

My blogging strategy acknowledged that some of that knowledge was not worth protecting. AI has now made most of it freely available.

This is not theoretical. I have watched it happen in my own series. Last week I listed the projects I built without a developer: contract routing, the Outlook-Todoist integration, the career development system. Those same projects are evidence against consultants, too. None of them required someone with more knowledge than me, plus AI. The knowledge asymmetry that would have justified a consulting engagement simply did not exist.

And it is not just me. Our own consultants now use the Microsoft Learn MCP server to look up base product functionality during client engagements. That is not a criticism of my team. It is an honest acknowledgment that no human carries the complete product knowledge base in their head. They never did. The difference is that the tool filling those gaps is now available to our clients, not just to us. The knowledge advantage that justified our rates is increasingly shared.

A client with AI can now get training without a trainer, answers without a support desk, implementation guidance without an implementation partner, and best practices without a best practices consultant. They can get all of this at any hour, with unlimited patience, with no billable clock running, and in the context of their specific data, their specific configuration, their specific business.

If consulting is the transfer of knowledge from someone who has it to someone who needs it, then the case against consultants is straightforward. The knowledge transfer mechanism has been replaced by something faster, cheaper, and always available.

Why consultants give wrong advice with confidence

The case against consultants does not stop at what clients can now do without us. There is a deeper problem, and I have already confessed to it.

In Week 31, I wrote about discovering that I had spent years giving clients the “Microsoft answer,” recommending that everything be built inside Business Central, not because I had rigorously evaluated the alternatives, but because the habit of confidence had outlasted the conditions that justified it. The world had shifted toward SaaS, open APIs, and integration platforms, making my default recommendation incomplete at best. Nothing in my professional environment, not partner incentives, not co-sell revenue targets, not peer behavior, flagged the drift.

I borrowed the philosopher Harry Frankfurt’s framework: I was not lying, I was bullshitting. I was producing confident output without having done the work to verify whether my frame was still accurate.

And I am not unique. This is a structural feature of consulting, not a personal failing. Every consultant operates within a framework shaped by their expertise, incentive structure, vendor relationships, and accumulated experience. That frame produces confident advice, and confident advice feels like expertise to both the client and the consultant. But the confidence of the delivery tells you nothing about the accuracy of the content.

We have a word for when AI does this. We call it hallucination, and we treat it as a disqualifying flaw. When a consultant does the same thing, giving confident advice from within an unexamined frame, we call it experience. The double standard is remarkable once you see it.

AI actually has an advantage here. You can verify AI’s output against source data. You can run the same question through multiple models and compare. You can ask AI to argue against its own recommendation, or to identify what it might be missing and where its reasoning is weakest. Try asking your consultant to do that. Try asking any trusted advisor to systematically argue against their own recommendation, identify their blind spots, and flag where their incentive structure might be shaping their advice. It is not that they will not, it is that the professional context makes it almost impossible. You do not get rewarded for undermining your own credibility.

Consultant hallucinations are not as real as AI hallucinations; they are harder to detect. There is no verification protocol. The client has no way to know whether the consultant’s recommendation reflects the full landscape or just the corner of it that the consultant happens to occupy. And the consultant often does not know either. I did not, for years.

The solutioning trap: when consultants skip diagnosis

There is one more piece of this, and it hits closest to home.

In Week 24, a client came to us wanting to automate their quoting process. They had done their homework and even used AI to generate a detailed requirements and solution document. Our consultant’s instinct was to start building toward the client’s specification before we had even signed a statement of work.

I had to pump the brakes. Our job is not to build what the client asks for. The first step is to determine whether what they are asking for is actually what they need.

But here is what makes this uncomfortable for the consulting steelman: the client had already used AI to do the diagnostic work we claim as our value proposition. They had analyzed their process, identified the bottleneck, researched approaches, and specified a solution. Our consultant’s instinct was not to challenge the diagnosis; it was to execute the prescription.

If the consultant’s reflex is to build what the client specifies rather than question whether the specification is right, and the client is increasingly capable of generating that specification with AI, what exactly is the consultant adding? The diagnostic judgment that is supposed to differentiate us from execution, the thing I had to remind my own consultant to do, is not being consistently delivered.

That is not an AI problem; it is a consulting problem that existed before AI. But AI makes it visible, because now the client shows up with work product that used to be the consultant’s deliverable, and the consultant’s response reveals how well they were doing the diagnostic work in the first place.

Where the case against consultants breaks down

So let me state the steelman plainly.

Clients can now access the knowledge that consultants used to keep behind billable hours. The knowledge transfer model has been replaced by something faster and cheaper. Consultants operate inside frames that produce confident but sometimes inaccurate advice, and unlike AI, there is no built-in verification mechanism. The diagnostic judgment intended to distinguish consulting from execution is not consistently applied, even within our own team.

If I am being honest, that is a strong case, stronger than the case against developers, because the walls are less concrete. A developer can point to the single-user boundary, the compilation step, and the production deployment. A consultant’s value is more intangible, and intangible value is harder to defend when the client is holding an AI-generated requirements document that looks an awful lot like what they used to pay you to produce.

Now let me tell you what that case misses.

Why AI does not replace accountability

Here is something 36 weeks of intensive AI use has taught me that I did not expect: the more AI helps me accomplish, the more ideas I have, and the more goes undone.

AI has not given me more hours in my day. It has expanded what is possible within the same hours. I can see more opportunities, generate more solutions, and explore more approaches than I ever could before. Every one of those ideas requires my time, my thought, my energy, and my judgment to evaluate, prioritize, and execute. The backlog of things I would love to accomplish grows faster than my capacity to work through it.

AI is like working with an extraordinarily capable but eager intern. It is willing to do almost anything I ask, and able to produce work product at a speed that would have been unimaginable two years ago. But the minute I stop directing, it stops. It does not follow up. It does not notice that I have not worked on the project for two weeks. It does not push back when I keep deferring the hard decision. It does not say, “You told me this was a priority last month and you have not acted on it.” It waits, perfectly patient and perfectly passive, for me to tell it what to do next.

Yes, I can set a reminder. I can build an automated follow-up system. I can set calendar alerts to remind me to revisit the project I have been avoiding. I can also ignore every one of those reminders, and I do, regularly, because a notification does not care whether I act on it. No one on the other end notices that I have dismissed it for the third time this week. A reminder is information. What I actually need is someone who will not let me off the hook.

Some people will say that it is the realm of agentic AI, systems that act autonomously and do not wait for direction. But that would require AI to know what I want to accomplish before I do, to understand my priorities, my constraints, my competing commitments, and the trade-offs I am not yet ready to articulate. The most important work in my day is not rote tasks waiting to be automated. It is judgment calls, strategic thinking, and creative problem-solving that I cannot delegate to an autonomous system, because I have not finished thinking them through yet.

Our clients face this same problem, magnified. They are not sitting around waiting for consulting insights they could generate themselves with AI. They are trying to fit the work we do together into small pockets of time carved out between their actual jobs. They are busy beyond belief and only getting busier. We are constantly fighting to get their time and attention to accomplish the things they have already paid us to help them do. AI does not solve this; it makes it worse because now they have even more ideas, even more possibilities, and even less time to pursue any of them.

Why consultants are like trainers, not textbooks

I work with a personal trainer, not because I lack access to workout knowledge. Every exercise program ever designed is available for free online, and AI can generate a perfectly customized training plan based on my goals, body type, and schedule in about 30 seconds.

I use a trainer because I need an accountability partner. Someone who shows up when every other part of my life is conspiring against regular exercise. Someone who notices when my form is slipping, adjusts the plan when a travel schedule or an injury changes the constraints, pushes me past what I would choose for myself, and holds me to the standard I said I wanted to achieve.

The trainer does not exercise for me. The results require my effort, my consistency, my willingness to show up and do the work. But without the trainer, I already know what happens: I skip sessions, I take shortcuts, I let the urgent crowd out the important. The knowledge was never the bottleneck. The sustained attention was.

Consulting at its best works the same way.

A client can use AI to generate an implementation plan. They can research best practices, analyze their processes, and design their future state. But someone still needs to show up every week and drive progress. Someone needs to notice when the project is drifting, when the finance team is quietly disengaging, when the warehouse manager is building workarounds instead of adopting the new process, when the executive sponsor is losing patience but not saying so. Someone needs to hold the client accountable to the timeline they committed to and push back when they want to defer hard decisions for the third meeting in a row.

AI cannot do that. Not because it lacks intelligence, but because sustained attention applied to a specific situation over time, with the relational context to know when to push and when to back off, is fundamentally a human capability. AI does not have a relationship with your CFO. It does not know that your warehouse manager agreed to the new process in the meeting and went back to doing things the old way the next morning. It does not recognize the political dynamics silently killing your project.

A real example: AI-generated specs are not enough

Remember the quoting engagement from Week 24, the one where our consultant started building to the client’s AI-generated spec before we had validated the approach?

Here is what happened after I pumped the brakes. We went back and reviewed the client’s actual process with them. What we found was what you almost always find when you compare a documented process to the real one: missing information, edge cases nobody had mentioned, and context that had not made it into the AI-generated requirements.

But the more important discovery was about the solution itself. The approach AI generated was not the best. It was heavily influenced by how the client had prompted AI, which makes sense once you think about it. AI is excellent at generating solutions based on what you ask for, but it is not great at helping you consider options outside the context you provide. The client described their problem through a particular lens, AI designed a solution through that same lens, and the result was a detailed, logical plan built on an incomplete picture. Not because AI failed, but because AI worked exactly as designed, within the boundaries of what it was given.

That is the diagnostic work: not answering questions, but seeing what is missing from the questions being asked. Not processing the client’s description of their problem, but recognizing that the description itself is shaped by the client’s frame, and that AI inherits that frame rather than challenging it. AI is brilliant at processing what you give it. It cannot walk through a warehouse and notice that the receiving process does not match what is documented in the system. It cannot sit in a meeting and read the body language that tells you the operations VP agreed to something she has no intention of implementing.

Why great AI analysis still fails to persuade

There is one more dimension to this that goes beyond diagnosis and accountability.

AI has helped me produce some of the best research, analysis, and visualizations I have ever created: sharper data, cleaner presentations, more thorough comparisons, and more compelling evidence. And I am still struggling to convince people who need to be convinced.

I wrote about this in Week 30: knowledge is power, but power is not persuasion. What I have learned since then is that this is not just true at home with my wife and daughter. It is true in every professional engagement. People do not make decisions based solely on the facts, no matter how well-presented those facts are. They make decisions emotionally and justify those decisions with curated facts that support positions they have already taken. They protect their turf, defer to relationships over data, and resist conclusions that threaten their comfort or their status, even when the analysis is airtight.

This is where the consulting value goes beyond anything AI can touch. AI can produce the perfect analysis. The consultant’s job is to navigate the room where that analysis lands: to understand who is resistant and why, to know when to push and when to let someone arrive at the conclusion themselves, to read the dynamics that determine whether a beautifully supported recommendation gets implemented or gets filed away and forgotten.

No amount of better data solves a stakeholder who has already decided. That takes a human who understands how humans actually work.

The real value of consultants in the AI era

The consulting model that is dying deserved to die. If you are selling access to knowledge, AI has replaced you. If you are selling confident answers from inside an unexamined frame, AI has replaced you and might actually be more reliable. If your instinct when a client hands you a specification is to build to it rather than challenge it, AI can execute that function faster and cheaper.

But if consulting is sustained attention, showing up consistently, driving accountability, diagnosing what the client cannot see because they are inside it, navigating the human dynamics that determine whether a project succeeds or fails, and pushing past “good enough” toward what the client actually needs, then AI has not replaced that. AI has made it more valuable because the gap between what is possible and what clients can execute on their own has never been wider.

The irony is this: AI has made clients more capable than ever at generating ideas, analyzing problems, and designing solutions, and that capability has made the consultant’s role more important, not less, because someone still needs to help clients navigate from “I know what I should do” to “I’ve actually done it.” The trainer does not become obsolete when better workout information becomes available. The trainer becomes more valuable because the client now has higher expectations and needs more sophisticated guidance to reach them.

The consultants who will thrive are not the ones defending the old knowledge-transfer model. They are the ones who have already moved past it, showing up as accountability partners, diagnostic thinkers, and agents of sustained attention in a world where their clients are drowning in capability but starving for capacity.

Self-audit: Is your consulting role at risk from AI

1. If you are a consultant, divide your last month of client work into three categories. How much knowledge transfer could a client have gotten from AI? How much was diagnostic work, uncovering problems the client did not know they had? How much was sustained attention, driving accountability, managing stakeholders, and pushing past “good enough”? The ratio tells you how vulnerable your role is.
2. If you are a client, think about the last consulting engagement you paid for. What did you actually get that you could not have gotten from AI? If the honest answer is “not much,” that is worth a conversation with your partner. If the honest answer includes moments where someone pushed you past your own resistance, saw what you could not see, or held you accountable to your own commitments, that is the value worth paying for.
3. Try the trainer test. Pick something you have been using a consultant for and spend a week trying to accomplish it with AI instead. Notice where you hit the wall. Is it knowledge? Or is it attention, accountability, and the ability to see your own blind spots?

What is next in this AI and jobs series

This is the second in a series. Last blog series: Do we still need developers? Next blog series: Do we still need partners? And then: Should not AI make everything cheaper?

The answers build on one another, and the pricing question might be the most uncomfortable of all.

This post is part of my “AI Over 40” series. It first appeared on LinkedIn: Do We Still Need Consultants?

Read more AI and Copilot blogs.

Trending Posts