That, in one scene, is the entire debate. And the stakes aren't philosophical. They're liability and profitability — both can kill restoration companies.

Artificial Intelligence is a brilliant apprentice and a terrible journeyman. Know the difference before it costs you more than what you saved on training and cultivating an actual journeyman. – Howie’s Helpful Hint (“HHH”)

For this article, let’s begin with what defines AI and RI. AI, Artificial Intelligence, is typically deployed as a search and answer mechanism like an embedded chatbot in an app or website; a workflow or research enhancement mechanism like ChatGPT, Claude, Grok, or CoPilot; or, an emergence in common industry platforms as an enhancement like Encircle, Cotality, KnowHow, DocuSketch, Verisk/Xactimate/Xactanalysis, Mikey’s Board, Clean Claims, and an infinite number of others.

RI, Real Intelligence, is generated by human thought processes through training, knowledge, experience, and most importantly, wisdom. 

Give the Machine its Due Respect

I'm not here to play Luddite, nor will I wax poetic about the way we used to do it, as I am a fan and an adopter of AI to simplify my mundane tasks. Roughly 88% of businesses now report using AI in at least one regular function. Check your CRM, estimating, ROM building, or project management platform release notes. The tools are here. Some of them are very good.

Where AI genuinely earns its keep:

• Workflow Development – It will build you a sequence — mobilization, containment, extraction, demo, drying, verification, reconstruction. It will not forget step nine because it was tired or use the company gas card to buy Cokes and Smokes. Consistency is a feature.
• Task coordination, communication, and execution – Automated assignments, escalation triggers, daily documentation reminders, and a paper trail that doesn't depend on remembering the information until the next morning, forgetting it, then pencil whipping it, and uploading it before the KPI dashboard changes color and SWAT shows up on your front lawn. All so you can be a good little Prole. Because if you don’t, money is held, rumors are started, and the “Party”  straps you to the table with giant diodes on your head and keeps repeating “Two plus two is Five.” Went all in on the Orwell 1984 references I suppose. If you still don’t get the reference, read more books and less Instagram posts.
• Scope development and engineering calculations – Dehumidification sizing, airflow and air changes, pressure differentials and containment, generator load and distribution — this is arithmetic with rules, and machines do arithmetic with rules better than we do. AI is infinitely faster, and without the fudge factor, hoping nobody notices, so you can rake on that extra $.07 to the bottom line you added.
• Reporting and data compilation – Psychrometric logs, moisture mapping, photo organization, daily reports assembled into something an adjuster can actually read without a field technician-to-administrator-to-estimator-to-adjuster translation app. The tedious, margin-eating documentation burden is exactly where automation belongs.
• ROM, estimate, and invoice creation – These platforms have been out for a while and the addition of AI to the common platforms have SUPER-charged them. However, this one is a double-edged sword. It is great to open an app, throw some figures into it and the world becomes a better place, rainbows become a superhighway to Happy Land, and your ROM, estimate, or invoice are perfect every time. The other side of the equation: lack of training, greed, and stupidity all raise their ugly head, as your work becomes perfectly wrong. My grandfather once told me that just because I was able to make his excavating equipment go, didn’t make me a heavy equipment operator.

Notice what every one of the bullets have in common: defined inputs, defined rules, defined outputs. Feed it a bounded problem and AI is the best technician you ever hired. It shows up at 3 a.m., never complains, and doesn't need a per diem. However, if the information input is by someone without training, knowledge, and experience, then disaster is inevitable. That someone then is nothing more than a booger-picking temp with a phone.

What Real Intelligence Still Owns Outright

Now walk AI onto a loss site. Here’s what AI cannot do…yet:

• Wisdom — Not knowledge — wisdom. There is a difference, and it's the load-bearing wall of this whole argument. AI will never be able to supplant wisdom because it will never have skin in the game. AI can accumulate knowledge at a rate no human can match. It cannot accumulate wisdom, because wisdom is what you get from being wrong in front of people and paying for it. Every experienced restorer in this industry privately carries their D.A.T. Catalog (Dumb Ass Things) full of mistakes, miscommunication, and trusting without verifying: the wall we didn't open, upholstery we ruined, the appliances we dropped, and the woodwork we destroyed , the “dry” reading we believed, the change order we didn't paper. Remember kids, a vase becomes a vâsse after the new guy drops it. The D.A.T. Catalog is why we're careful because we have skin in the game. A model or platform has no catalog. It has never eaten a $400,000 mold claim and lain awake over it.
• Situational awareness in decision-making – The 40-year-old building where the as-built drawings are creative fiction. The pungent yet intoxicating smell that clues you in that there may be an issue behind that wall. The tenant who is technically not your client but will absolutely become your problem by Thursday. These are examples of gray area variables. There is no effective way to create a rule for AI. It requires RI to have that gut feeling to pivot as needed to keep moving forward.
• Reading and shaping the room – Impression, presence, credibility. A homeowner standing in six inches of water at midnight is not evaluating your psychrometric prowess and cares even less about counting grains of moisture in the air. They're deciding whether to trust you. That decision gets made in about eleven seconds, the same eleven seconds the chatbot used to be confidently wrong. The point is your credibility is based on human signals, not moisture logs.
• Empathy, and knowing which emotion the moment calls for –  Knowing how to read the room, then proactively and professionally being the person you need to be without changing your values. Calm for the panicked homeowner. Precision for the adjuster. Blunt for the consultant who's stalling. Warmth at the kitchen table. Cold steel in a claim settlement meeting. Same person, same day, different register, because you read the situation.
• Aligning interests and perceptions – Half of what we do is getting an owner, a carrier, a TPA, a hygienist, and a property manager pointed the same direction. That is negotiation, and negotiation is a contact sport.
• Reflexive adjustment on incidental discovery – You open a wall and find asbestos-suspect mastic, a live circuit where none was drawn, or a second loss nobody reported. RI stops, re-thinks, re-scopes, and re-prices. AI keeps executing the plan, unless you stop and ask it. And even then, the response will be confusing or less than what you really need to know. When there are too many variables to consider, or if there is any gray area in decision making, the chatbot is rendered virtually useless.
• Brand at the personal level – Nobody ever referred a contractor because their chatbot on their website was polite.

Where Each One Breaks

AI's failure modes are dangerous precisely because they're fluent — It is confidently wrong. It fills gaps with plausible fiction. It doesn't know what it wasn't told; and this is the killer: it never announces that it's out of its depth. It also has no standing, no license, no certification, and no professional liability. It has no skin in the game.

When an AI-generated scope misses a critical item and the claim goes sideways, the Errors & Omissions carrier is not going to accept “the software said so.” You signed it. You own it. Garbage in, garbage out — except now the garbage comes back beautifully formatted with a confidence level attached. — HHH

RI's failure modes are the ones we've always had – We're inconsistent. We're tired at hour fourteen. We do arithmetic in our heads when we shouldn't. We document under duress and it shows. We're expensive, we don't scale, and we take our institutional knowledge home with us when we retire, and to the grave after. Given who owns most of the restoration companies in this country, as over 70% of owners are Baby Boomer or early Gen X, it’s a problem arriving faster than anybody's succession plan can solve.

The trend that should scare you

Here's what's happening in the market, and it isn't AI writing better scopes.

A survey of 1,000 U.S. business leaders found that 37% expect to have replaced jobs with AI by the end of 2026, with recently hired and entry-level employees at the highest risk. Read that again. Entry-level. The trainees. The people who were supposed to become your next generation of project managers by making supervised mistakes on real jobs.

Companies are quietly swapping mentorship for a help desk. Instead of a foreman walking a tech through why we're pulling that baseboard, the tech gets a chatbot answer. It's cheap, it's instant, and it feels like training. It isn't. It's lookup. You're producing people who can retrieve an answer and cannot evaluate one — which is the most expensive kind of employee you can put on a commercial loss.

The corporate world is already walking this back. Klarna famously replaced roughly 700 customer service agents with AI, then reversed course after its CEO conceded the output was “lower quality” and started hiring humans again. A MIT report last year found that the overwhelming majority of enterprise generative AI pilots delivered no measurable return, the number's been argued over, but the mechanism isn't: firms bought tools to handle AI recommendations instead of building capability and capacity.

We are a trade that runs on judgment formed at 2 a.m. in a wet building. That judgement has always been formed out of training, knowledge and experience. If we hollow out the apprenticeship, the training, and creating a corporate culture of knowledge and safety to save payroll, we will not feel it this month or this quarter. In fact, we will feel confident we made the right decision. But, and it’s a big but, we'll feel it in about four to six years, when there's nobody left who knows why the answer on the screen is wrong.

Harmonizing the two: draw the line at consequence

The line isn't hard to figure out,  Automate the repeatable. Never automate the consequential.

Let AI carry the load it's built for: calculations, logs, documentation, scheduling, estimate reconciliation, first-draft scopes, invoice assembly. Then require that every output crosses a qualified human desk for accuracy and quality control before it goes to a client, a carrier, or a courtroom. Put a name on it. That's not bureaucracy — that's your liability firewall, and it costs you minutes versus your legacy.

Take the hours AI just gave you back and spend them on the things that will compound return on your security, your wealth and your corporate legacy. Spend field time with your people, real mentorship, root-cause reviews of jobs that went sideways, and capturing what your veterans know before they retire to compound return on the base of knowledge in your company that AI can use in the future. AI may hold institutional knowledge like debriefs, lessons learned, “here's why we did it that way”, but it is the human resources who create it. Machines are excellent librarians and lousy authors. 

Whether you’re an owner looking to truly improve your times x on your valuation or a tech, administrator, or manager looking to increase your market value in the industry, remember being the one who can solve problems and become the author of  institutional knowledge is more valuable than the booger picker who looks up the answer on a chatbot and still gets it wrong. – HHH 

It is the new profit equation. Automation through AI drops your cost per project. Wisdom drops your reservicing rate, callbacks, supplements, and liability claims filed against you. The first improves the quarter. The latter makes the company or individual an invaluable asset as a problem solver.

A Call to Action

We're all juggling the same balls: what the customer needs, what the building requires, what the carrier will pay, what the crew can execute, and what the calendar allows. AI adds capacity to that juggle. It does not add judgment.

Audit where you're using and relying on AI based platforms and be honest about which side of the consequence line each use falls on. Insist on a human signature on anything that creates accountability and improves quality. Rebuild the apprenticeship you've been tempted to automate away. Pair every tool with a mentor. Debrief your losses on purpose and write down what you learned, while the people who learned it are still on the payroll. Measure the right things: not just cycle time, but reservice rate, supplement rate, and whether your people are getting better or just getting faster.

Do that and everybody wins. This is a business where you meet the best people, companies or causes at the worst time. The competent problem solver who properly uses AI as a tool and not a replacement for wisdom is the true future of the industry. The customer gets back in the building sooner. The employee gets a career instead of a script. The employer gets margin, legacy, and wealth instead of rework, fantasy financial pictures, and liability that no insurance can cover. The industry keeps the hard-won practical knowledge that no dataset will ever contain — because you can't train an AI model on regret.

AI will make your company smarter. Only your people can make it wise and consistently profitable. – HHH