Remodel, Gut Rehab, or New Construction: Three Ways Enterprises Build for AI

Merivant — 2026

Walk the Jobsite First

Every enterprise facing AI transformation eventually arrives at the same question. Not “which model?” or “which vendor?” The question is the one every contractor asks before giving you a number: what are we working with, and how much of it are we keeping?

The answer determines everything. Budget, timeline, risk, how much of the operation shuts down during construction, and whether people can keep working while the walls come down. Get this wrong and you are six months into a remodel that should have been a gut rehab, or breaking ground on new construction when the foundation you already have is perfectly sound.

Construction figured this out centuries ago. Three types of project. Three risk profiles. Three honest conversations about what stays and what goes. Enterprise AI transformation has the same three — it just hasn’t been using the right language.

Remodel

What it is: The building stays. The bones are good. You update the fixtures, rewire a few circuits, knock out a non-load-bearing wall to open the floor plan. AI enters as a renovation — targeted, bounded, livable during construction.

The jobsite: A claims processing team running a 14-day cycle gets an AI triage layer that cuts it to 3. A support desk drowning in L1 tickets gets an agent that resolves 40% before a human touches them. A finance team reconciling spreadsheets by hand gets an extraction pipeline. Each project has a scope, a punch list, and a certificate of occupancy. You live in the house the whole time.

Who hires this contractor: Regulated industries. Healthcare systems that cannot afford to close the building. Financial institutions where the compliance framework took a decade to pass inspection. Any organization where the cost of shutting down exceeds the cost of dated finishes.

The real advantage: Speed to occupancy. Over 74% of executives using focused agentic AI deployments see returns in the first year. You are not asking anyone to move out. You are asking them to live through some dust and noise in exchange for a better kitchen. Teams adopt because their daily work gets easier, not different.

What you find behind the wall: Remodels have a universal risk — you open up a wall expecting to run new wire and discover the plumbing is rotted, the framing is wrong, or there is no insulation. In enterprise terms: you instrument a process with AI and discover the process itself is broken, the data is garbage, or the upstream system produces outputs no model can reason over. Every remodel carries the risk of revealing that you needed a gut rehab all along.

The trap: Remodeling optimizes the current floor plan. If the floor plan is wrong — if the rooms are in the wrong places, if the building faces the wrong direction — you end up with a beautifully renovated version of a layout that does not work. Nearly a third of CIOs now prefer this “supplement, don’t replace” approach. That is the safe bid. Sometimes the safe bid is how you spend $200,000 improving a building that should have been torn down.

When it fails: When the foundation is the problem. When the data is siloed so deeply that no amount of point solutions can reach across load-bearing walls. When leadership uses “remodel” as a euphemism for “we are not actually willing to move anything.”

Signs you hired the right crew: Each room delivers standalone value. The next project costs less than the last because you learned where the pipes run. Teams stop calling it “the AI project” and start calling it “how we work now.”

Gut Rehab

What it is: Strip to the studs. Keep the foundation, the shell, the address. Everything inside comes out — plumbing, electrical, HVAC, layout. When it is done, it looks like the same building from the street. Inside, nothing is the same.

The jobsite: An insurance company does not rewrite its 30-year-old policy administration system. It does not slap chatbots on the customer portal either. It strips the interior: builds a modern event-driven layer that reads from the legacy system, enriches with AI — classification, risk scoring, document extraction — and exposes clean APIs that new frontends and agents consume. The legacy system still runs. But it is no longer the living space. It is the foundation beneath a completely redesigned interior.

A manufacturer does not replace its ERP. It guts the integration layer between ERP, shop floor, and customer systems, rebuilds it with an AI-native data fabric, and lets the old ERP serve as a system of record behind a smarter boundary. The walls moved. The wiring is new. The address is the same.

Who hires this contractor: Mid-market companies and large enterprises with buildings worth saving. Organizations with solid foundations — market position, customer relationships, institutional knowledge, brand — but interiors that no longer serve how the work actually happens. Teams with strong project management discipline, because a gut rehab without a schedule becomes a money pit.

The real advantage: Decoupling. The old plumbing does not need to know about the new HVAC. The new electrical does not need to follow the old wire paths. Generative AI accelerates this kind of modernization by 40–50% because the new interior creates clean surfaces for AI to operate against. New capabilities ship without touching the foundation. Old systems retire without disrupting new capabilities. You get a modern building on a proven lot.

The trap: Scope creep to new construction. Every gut rehab reaches a moment where someone says “as long as we have the walls open, we should also...” and suddenly the budget doubles. The discipline is knowing what the shell is worth. If you are keeping the foundation out of sentiment rather than structural soundness, you are paying gut-rehab prices for a building that should have been demolished.

The other trap: the project takes so long that you cannot occupy the building during construction, and the business operates out of a temporary space that quietly becomes permanent. Gut rehabs that stretch past their timeline create two systems — the old one everyone is still using and the new one nobody has moved into.

When it fails: When the foundation itself is cracked. When the shell is not worth the cost of saving it. When “gut rehab” becomes “permanent renovation” — a jobsite that never reaches move-in day. When governance gaps in the new interior create problems the old layout never had.

Signs you hired the right crew: New capabilities ship faster after the rehab than before it. Old systems start getting decommissioned, not just wrapped. The architect’s backlog comes from business demand, not internal ambition. People move back in and the building works better than the day it was built.

New Construction

What it is: Empty lot. New blueprints. Purpose-built for how the work happens now, not constrained by what someone framed 20 years ago. The old building is not renovated or gutted — it is demolished or abandoned. You are building from dirt.

The jobsite: A logistics company does not add AI routing to its existing dispatch system. It builds a new dispatch system where AI routing is the core loop and humans handle exceptions. A media company does not add AI writing tools to its CMS. It builds a content engine where AI generates, humans curate, and the CMS is a rendering layer, not the creative workspace. The old system is not migrated. It is replaced.

A startup has the simplest version of this: there is no old building. Just a lot and a set of blueprints designed for AI-native operations from the first pour of concrete.

Who hires this contractor: Organizations with condemned buildings — legacy systems that are actively failing, costing more to maintain than to replace, or architecturally incapable of supporting what the business needs next. Companies in existential crisis where the market has already moved to a new neighborhood. Startups building on vacant land.

The real advantage: No integration tax. No “the old wiring can’t handle that” conversations. No load-bearing walls where you need open space. The building is designed for current code, current loads, current occupancy patterns. AI is not retrofitted into an old floor plan — it is the floor plan.

The trap: New construction is where the money goes to die. The $30 to $40 billion invested globally in generative AI, with 90–95% of organizations seeing no measurable return — a disproportionate share went to new construction projects that never reached occupancy. Buildings that went over budget, over schedule, and over the patience of the people funding them.

New construction requires the most organizational change capacity. You are not asking people to tolerate dust. You are asking them to move to a different address, learn a new layout, and abandon the muscle memory they built in the old space. That change management cost is underestimated by an order of magnitude in most project plans.

When it fails: When the old building encoded institutional knowledge that nobody documented before demolition. When the timeline assumes the organization can operate without a building for 18 months and it cannot. When the blueprints are drawn by architects who never worked in the old space. When “new construction” is driven by the excitement of a blank page rather than the necessity of a condemned structure.

Signs you hired the right crew: The old building was already failing — you are not demolishing something functional. Maintenance costs on the old system exceeded replacement cost. The organization has survived a major move before and has the logistics muscle for it. Leadership is committed past the foundation pour, not just the groundbreaking ceremony.

The Inspection Report

Choosing is not about which project sounds best in the proposal. It is about what the inspection reveals. Three questions:

1. What does the building inspection say?

2. Can you stay in the building during construction?

3. What is worth more — the building or the lot?

Every Block Runs All Three

Most enterprises are not a single building. They are a campus. And most campuses run all three projects simultaneously. A remodel in finance where the processes work and just need AI acceleration. A gut rehab in operations where the data is valuable but the systems are failing. New construction in a greenfield division where there is nothing to renovate.

That is fine. Healthy, even. The mistake is not running multiple projects. The mistake is running them without knowing which one is which. When a remodel silently becomes a gut rehab because every wall you open reveals another problem, that is how budgets evaporate and timelines collapse. When a gut rehab is managed like a remodel — phased room by room with no overall blueprint — you get a building where nothing connects.

Name the project. Scope the project. Fund it for what it actually is. The GC who tells you “we’ll figure it out as we go” is the one who puts you over budget. The one who walks the jobsite, reads the inspection report, and says “here’s what we’re looking at” is the one who finishes.

Breaking Ground

If you chose Remodel: Pick three rooms. Measure them now — cycle time, error rate, manual hours. Instrument with AI. Measure again in 90 days. If the numbers moved, open the next room. If they did not, the room was not the problem. Pull permits for an inspection before you swing again.

If you chose Gut Rehab: Start with the lot survey, not the blueprints. What data do your AI capabilities actually need? Where does it live? What condition is it in? Design the new interior around the data that matters, not an abstract floor plan. A gut rehab that starts from occupant needs delivers. One that starts from architect ambitions stalls.

If you chose New Construction: Staff it with people who lived in the old building, not just people who will build the new one. The institutional knowledge in the heads of long-term occupants is the most valuable input to the new design. Lose it and you will build the same floor plan with newer materials.

Regardless of project type: inspections are not a phase. Code compliance runs parallel from day one. Most construction failures come from structural and permitting gaps, not bad materials. The AI works. The question is whether your organization can occupy what the AI makes possible.

Merivant builds AI systems that fit inside human behavior, not the other way around. merivant.com