Most AEC Firms Stall on AI Because Nobody Owns the Decision

The Accountability Vacuum

Most mid-size AEC firms don't fail at AI because they picked the wrong tool. They fail because no one was ever named accountable for the decision, and "the leadership team owns it" quietly becomes a way to avoid naming anyone. It sounds like consensus. It functions like nobody's job.

Fixing that takes a 90-day AI ownership charter for a mid-size AEC firm: one page, four parts, done before the next quarter starts. Not a policy binder. Not a committee that meets forever without deciding anything— it's an AI strategy built around who actually decides.

Companies everywhere are deploying AI at production speed while governing it at committee speed¹. Spreading responsibility across IT, ops, and whichever principal champions a pilot "obscures rather than clarifies who ultimately answers when failures occur."¹ Zweig Group, the AEC-specific consultancy, puts it plainly: assign clear ownership rather than dispersing responsibility².

Naming one person out loud, in a room full of partners, feels riskier than it is— which is exactly why "the whole leadership team owns it" survives as the easier thing to say. It's a governance answer that avoids the governance question. A named AI owner is where this starts, and it's why the rest of this charter exists at all.

Why AEC's Liability Exposure Changes the Calculus

AEC firms carry a liability exposure most industries don't. Stamped drawings and sign-and-seal contracts put a named professional's license behind AI-assisted work, and that's exactly why an unnamed "everyone" can't be the owner.

Zweig Group frames AI adoption around trust rather than raw technology². "AI isn't a tech problem – it's a trust problem," their April 2026 analysis puts it²— and the tension shows up in five places specific to how AEC firms actually work:

• Innovation vs. Liability— the upside of a faster deliverable against the downside of a stamped error
• Speed vs. Review Capacity— AI produces work faster than a principal can review and seal it
• General Tools vs. Purpose-Built— ChatGPT versus AEC-specific software, and who decides which gets used where
• Firm-Wide vs. Project-Specific— one AI policy for the whole firm, or one per contract
• Transparency vs. Fear— telling clients AI touched their deliverable, or staying quiet about it

The named-owner vs. committee finding isn't unique to AEC. Federal guidance and research from hospitals and banks land in the same place— which tells you this is a governance problem, not an industry quirk. What's different is the stakes and the staffing: a mid-size firm rarely has the CISO or General Counsel a generic template assumes, and a stamped drawing carries a license a marketing deck never will. As Zweig Group puts it: "In AEC, we do not move fast and break things. We sign and seal them."²

Element 1: Name a Single Accountable Owner

The first line of the charter is a name, not a department: one person, the named AI owner, holds end-to-end accountability for AI decisions, even when a cross-functional group advises on them.

RACI (Responsible, Accountable, Consulted, Informed) is the framework doing the work here, and its rule for AI decisions is non-negotiable: exactly one Accountable owner. Everyone else is Responsible, Consulted, or Informed— never Accountable³. NIST, the federal agency behind the AI Risk Management Framework most large organizations already reference, backs this up: "accountability structures" must be "in place so that the appropriate teams and individuals are empowered, responsible, and trained" to manage AI risk⁴.

In practical terms: you're looking for the person who will still be answering for the AI decision six months after everyone else has moved on.

Do you need a full-time Chief AI Officer to fill that Accountable box? No. An existing role can absorb it: a COO, a CTO, or a discipline principal. Whoever already owns AI risk today, whether they've said so out loud or not, as long as that ownership gets written down instead of left implicit. That written-down part is the whole game. 76% of organizations now have a Chief AI Officer, up from 26% a year ago⁵— momentum toward naming someone, not proof a dedicated hire is required. IBM surveyed CEOs at large enterprises for that number, not 100-500 person AEC firms, so read it as a signal of where the practice is headed, not a mandate for your firm's size. Weighing a dedicated hire against handing this to someone already on staff? A fractional AI officer is worth reading first.

Ownership on a charter isn't real until the name on it has actual skin in the game. A title assigned in a memo is not the same thing as someone who will answer for the outcome, and if you can't picture the specific person nodding when the charter gets read aloud in a partner meeting, you haven't finished naming the owner yet.

Element 2: Define What the Committee Actually Does

An AI governance committee charter names exactly one Accountable owner under RACI³. And the committee's job is to advise that owner— not to replace them. Its output is a recommendation, not a decision.

Naming one owner doesn't disband the committee. It gives the committee a clear job: coordinate input, then hand a recommendation to the person who decides. A structural engineer, an architect, and an MEP lead see different AI risk in the same tool, and an owner who never hears from any of them is flying blind. Tunnel vision is real when one person decides alone— which is exactly why the committee keeps coordinating while one person stays accountable for the call, per Reconn's RACI model and NIST's Govern function³⁴.

Reconn's own template builds the committee around an enterprise roster: a CTO as Chair, plus a Chief Risk Officer, General Counsel, CISO, and Data Protection Officer³— titles most 100-500 person AEC firms don't have. Right-sized, it's closer to: the owner as chair, one discipline lead per service line, an IT or ops lead, and whoever already handles risk and insurance. Every meeting, the committee brings the owner:

• Which pilots are running, and where they stand
• Any new AI use case a discipline wants to try
• Anything that needs escalating before it becomes a problem
• Open policy questions the committee can't resolve itself

A committee that meets on schedule still needs one name attached to the decision, or the schedule is all that's accountable. ACEC, the American Council of Engineering Companies, names "Lack of Technology Leadership" as engineering firms' most commonly cited adoption barrier⁶; this structure answers that directly. See a clear AI decision framework for what a committee's recommendation should contain, and our guide to AI governance strategy for the policy layer this charter sits inside.

Element 3: Set the Meeting Cadence

Without a standing meeting, the charter becomes a document that gets signed once and drifts. Meet monthly while you're actively piloting AI tools, and no less than quarterly once deployments stabilize³.

This isn't a technology rule. Every governance structure has the same failure mode: accountability doesn't survive without a standing meeting on the calendar, no matter how good the project management tool underneath it is. A meeting that doesn't happen is a decision that doesn't get made. In AEC, a decision that doesn't get made is usually a pilot running without review— and a stamped deliverable waiting for the outcome.

Each meeting should produce three things, in writing: a decision log entry, a current list of open pilots, and any escalations that need the owner's attention before the next meeting.

Meeting cadence and the evidence trail matter more than headcount.

None of this runs parallel to Element 1— it's what feeds the owner's decisions. The committee talks monthly or quarterly; the owner decides continuously, using what the committee surfaces. Skip the meeting a few times, and the charter quietly reverts to exactly what it was built to fix: an accountability vacuum with a nicer name.

Element 4: What 90 Days Actually Buys You

In 90 days, the deliverable is a signed charter and a named owner in place before the next quarter starts— not firm-wide AI fluency.

Here's the checklist a firm should be able to check off by day 90:

• A named owner, written into the charter by title and name
• Decision rights and an escalation path, written down rather than assumed
• A meeting cadence already on the calendar for quarter two
• A defined success metric for quarter two, not just "keep going"

The governance gap in most AEC firms isn't about AI readiness— it's about who answers when a stamped deliverable involves AI-assisted work and something goes wrong.

The stakes for getting this right are real. MIT's NANDA research program found that roughly 95% of generative AI pilots at enterprises fail to produce measurable ROI, with integration and organizational gaps, not model quality, cited as the reason⁷. That study measured broad enterprise GenAI pilots, not AEC firms or governance structures specifically, so treat it as a fair counterweight rather than the final word: no owner has a documented failure pattern too.

Ninety days produces the document and the name on it. It doesn't produce a firm that's fluent in AI— that's a longer, separate project.

The Cost of Waiting

AI ownership structures are moving from optional to standard practice fast. 76% of organizations now have a Chief AI Officer, up from 26% a year ago⁵, and AEC firms report the same acceleration: 63% now say they have an AI strategy in place, up 11 points, and 78% believe AI will help their business, up 15 points— both figures from the same ACEC Research Institute survey program, not two separate confirmations⁶⁸.

Structure is arriving faster than results. That's exactly why naming an owner now costs less than naming one after a bad outcome forces the question.

This charter is the whole AI angle for a firm at this stage. Signing it doesn't make your firm AI-fluent, and it isn't meant to. But once the owner is named, something actually starts moving. It's a first concrete step, not the finish line— building AI culture after the charter is signed covers what comes next.

Drafting and right-sizing this charter without guessing is exactly where an implementation partner earns their first day.

The structure matters. But the answer to "who owns AI here" was never going to be a policy document. It's always a person. No matter the question, people are the answer.

FAQ

Who should own AI at an engineering or architecture firm?

One named individual should hold end-to-end accountability for AI decisions. A cross-functional group can still advise, weighing in on use cases and flagging risk, but shared ownership without a named owner functions as no ownership at all. If no single person answers for a decision, nobody does¹².

Does a mid-size firm need a full-time Chief AI Officer?

Not necessarily. An existing role, a COO, a CTO, or a discipline principal, can hold the accountability, as long as it's explicitly named and written into a charter. More organizations are creating a dedicated Chief AI Officer role, but that's a signal of where the practice is heading, not a requirement for a firm this size³⁵.

How often should an AI governance committee meet?

Monthly while the firm is actively piloting AI tools, and no less than quarterly once deployments stabilize. What matters more than the exact interval is that the meeting produces something in writing every time it happens: a decision log, a list of open pilots, any escalations³.

Why does AI governance matter more for AEC firms than other industries?

AEC firms sign and seal deliverables under professional licensure, which raises the liability stakes of AI-assisted work above those of a typical enterprise. A named professional's license sits behind the work in a way it usually doesn't in a generic corporate AI rollout².

What should be in a 90-day AI ownership charter?

A named owner, defined decision rights and an escalation path, a set meeting cadence, and stated success criteria for the following quarter. Those four elements are the deliverable— not a fully AI-fluent organization, which takes longer than a quarter³⁴.

References

1. CIO.com (Foundry), Pat Brans, "AI is spreading decision-making, but not accountability" (2026)— https://www.cio.com/article/4160986/ai-is-spreading-decision-making-but-not-accountability.html
2. Zweig Group, Julia Moroney, "AI governance and trust in AEC firms" (2026)— https://zweiggroup.com/blogs/the-zweig-letter/ai-governance-and-trust-in-aec-firms
3. Reconn, Shenoy Sandeep, "AI Governance Committee Charter: Template + RACI Guide" (2026)— https://orbit.reconn.io/ai-governance-committee-charter/
4. National Institute of Standards and Technology, "AI Risk Management Framework Playbook— Govern Function" (2023)— https://airc.nist.gov/airmf-resources/playbook/govern/
5. IBM Institute for Business Value, "IBM Study: CEOs Are Reshaping the C-Suite for the AI Era" (2026)— https://newsroom.ibm.com/2026-05-04-ibm-study-ceos-are-reshaping-c-suite-roles-for-the-ai-era
6. American Council of Engineering Companies (ACEC) Technology Committee, "Why and How?— A Primer on AI Integration for Engineering Firms" (2025)— https://www.acec.org/wp-content/uploads/2025/10/AI-Adoption-in-AEC-Summary_Primer_October_2025.pdf
7. MIT NANDA via Fortune, "MIT report: 95% of generative AI pilots at companies are failing" (2025)— https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/
8. ACEC Research Institute, "Engineering Firms Report Rebounding Economic Confidence, Accelerating AI Investment" (2026)— https://engineeringinc.acec.org/blog/engineering-firms-report-rebounding-economic-confidence-accelerating-ai-investment/