Does Enterprise Architecture Need A New Boss?

Enterprise architecture keeps getting handed a company-wide mandate without company-wide authority, and the reporting line to the CIO is usually blamed first. The sharper diagnosis is different: the problem isn't primarily who EA reports to, it's who actually backs it when two or more business units disagree, though reporting line and sponsorship quality are entangled enough that treating them as a clean either/or misses the point.

José Freitas is Lead Enterprise Architect at IATA and has spent fifteen-plus years moving between chief architect roles in banking, manufacturing, and M&A architecture work. He's blunt about where the friction sits.

"The authority to produce enterprise impact does not belong to a single organizational unit, but at a higher level, such as the CEO or the board," Freitas said. "When enterprise architecture sits inside the ICT business unit, its priorities default to an IT-centric point of view, not because IT is wrong, but because that's the agenda closest at hand. The board's priorities, and the rest of the business's, don't automatically get a seat in the room."

That risk is measurable, not just anecdotal. Industry surveys suggest a large majority of enterprise architects sit two or more levels below the CEO, with a minority reporting directly to the top. Separate research puts the perception gap in view: over 90% of CIOs describe EA as a strategic enabler, yet roughly half of enterprise architects say their insights go underused in actual strategic planning. That gap persists regardless of exact org-chart position, which is itself evidence that reporting line alone isn't the mechanism doing the damage. Something else is absorbing the difference between stated support and real influence.

• Reporting line is a symptom, not the disease: Some practitioners argue that C-suite engagement with chief architects holds up reasonably well one level below the CEO and only deteriorates further down the chain. That's plausible. Sponsorship typically erodes with distance from the top, but since no precise data exists on interaction quality by reporting depth, it should be treated as hypothesis, not fact. What the evidence does support, Freitas says, is the broader shape of the claim: the real fault line isn't CEO versus CIO. It's "close to the top with a strong sponsor" versus "buried where no one notices when EA loses a fight."

• The sponsorless advisor problem: An architect reporting to the CEO with no real sponsorship will likely lose a dispute to one reporting to a CIO with the CEO's active backing. Moving a box on the org chart is cosmetic. What changes outcomes is a standing mandate, a board or CEO decision that EA's findings carry weight in resource and priority conflicts, regardless of which unit raised the issue.

Freitas frames the function's job as translation rather than authorship of strategy. "People think you define the strategy. No. You understand the objectives defined by the board, and your obligation is to convey them using principles and other tools, so the business comes to its own strategy," he said. "We are stewards." EA isn't asking for decision rights over business strategy, but asking for a recognized voice when initiatives, capabilities, or investment claims collide.

• Where the bias actually shows up: "The CIO is a peer among peers, and in most organizations, although important, is a supporting function with no direct revenue attached, which doesn't help in negotiations," Freitas said. That framing holds in classic cost-center IT setups, though it applies less cleanly in organizations where the CIO holds a shared-services P&L or sits on the executive committee with real budget authority. The underlying point survives the nuance: without a mandate that outranks the CIO's peer-level standing, EA inherits the CIO's negotiating position, not the boards.

• Co-parenting someone else's child: Capability or value-stream ownership is where this becomes concrete and a personal. "It becomes like divorced parents fighting over custody," Freitas said. "People feel they're both responsible for it. It's like co-parenting, even though it's not really your kid. You feel it like it's yours, but in reality, it belongs to the company." Two business units can both claim ownership of a shared capability, or both disown it when something breaks, and without a named accountable owner and a defined escalation path, the capability model becomes a wall chart rather than a working tool. This pattern recurs across sectors consistently enough to treat it as a structural feature of immature business architecture, not a one-company quirk.

When EA's actual impact runs through business initiatives rather than IT delivery, binding its performance to ICT metrics (uptime, ticket volume, delivery against IT budget) measures the wrong thing. According to Freitas, swapping in outcome metrics without an attribution method just moves the problem downstream. In rough order of how defensible the attribution is:

• Capability-to-initiative traceability. The ratio of funded initiatives mapped to defined capabilities versus duplicated effort across units, the most directly attributable metric, since EA controls the mapping process itself.

• Rework and duplication avoided. A traceable proxy for value, estimable from documented decisions EA influenced before build started, and more defensible than a direct ROI claim.

• Time-to-market for EA-supported strategic initiatives, benchmarked against a matched set of comparable initiatives without EA involvement.

• Product and platform adoption rates following EA-guided roadmap decisions.

• Downstream customer experience movement (e.g., NPS) tied to EA-prioritized initiatives, reported strictly as correlation unless a controlled comparison exists.

The first two metrics should anchor any pilot, since attribution weakens progressively down the list. Before rolling this out organization-wide, Freitas advises piloting it on two or three initiatives, documenting how attribution was calculated, and let the board see the method before trusting the numbers. A KPI framework without a stated attribution method just replaces one set of misleading metrics with another.

• The missing referee: Even with better metrics, competing claims between business units need somewhere to go. "In many organizations, especially outside heavily regulated sectors, there's no defined escalation or arbitration path, so conflict gets suppressed or escalates informally to whoever holds the most political capital," Freitas explained. TOGAF's governance model anticipates this gap through the Architecture Review Board (ARB), giving it standing authority over architecture decisions and a structured process for setting the right sponsor level rather than mandating one fixed answer. In practice, TOGAF's own guidance points toward the highest practical level of sponsorship, and organizations that settle for less often find ARB findings stop at the CIO's desk rather than reaching whoever can break a tie between units.

The fix isn't a new layer. It's a fixed slot: either a dedicated ARB session with the board on a defined cadence, or a standing agenda item at recurring board meetings reserved for ARB escalations.

Board or CEO-adjacent positioning makes sense once an organization crosses a countable threshold. There's no universal formula. Freitas advises each organization to adjust these to its own risk tolerance and dispute frequency, then iterate to find the right balance:

• Three or more unresolved capability-ownership disputes escalated informally within a twelve-month period.

• A documented gap between business architecture maturity, scored against a standard maturity model, and the CIO's actual mandate scope.

• Regulatory exposure meeting a defined checklist: prudential regulation, safety-critical compliance, or cross-border data governance obligations.

Below that threshold, a CEO-sponsored advisory mandate for EA, still reporting to the CIO, with guaranteed ARB access to the board on escalation, solves most of the same problems with far less disruption. Structure and mandate aren't competing fixes. "The threshold model is really a decision rule for how much reporting-line change a given mandate needs to be credible," Freitas said. Neither works alone: a board-adjacent EA function with no real board attention is worse than a CIO-reporting one with strong sponsorship, because it creates the appearance of authority without the substance.

If this resonates, Freitas offers a direct way to test it: track CIO-reporting EA functions with guaranteed ARB access to the board over six months to a year, and compare capability-dispute resolution speed against a comparable group without that access. If sponsorship is the deciding variable, that gap should show up in the data, not just in interviews.

Freitas's closing point applies to governance design as much as talent: the scarcest resource isn't the reporting line, but people who can hold technical depth and stakeholder negotiation at once. "You need the industry knowledge and the negotiation skills," he said. "The clarity is for the stakeholders to make their best choices. We are just an enabler." A organization that fixes the mandate without investing in that capability, and without a way to verify it worked, has solved only half the problem.

The views and opinions expressed are those of José Freitas and do not represent the official policy or position of any organization.