Most portfolios don’t lack initiatives. They lack stopping power. Once funding is approved and a program is publicly endorsed, the system favors continued support over learning, even when evidence weakens. IT leaders are increasingly judged on whether they can prevent the enterprise from drifting into sunk cost inertia while protecting credibility. The ability to stop misaligned work without triggering political fallout has become a quiet leadership superpower.
That matters because the role itself has shifted. Research from McKinsey shows that nearly two-thirds of top-performing companies say their technology leaders are “very involved” in crafting enterprise strategy. Deloitte’s Tech Exec Survey reports that 80% of tech executives say their responsibilities have expanded significantly and that more than a third now manage a P&L. When technology leadership is part of enterprise strategy, portfolio stopping is not an IT hygiene issue. It is a business leadership capability.
I learned this the hard way in portfolio reviews that looked “healthy” on paper. Everything was green. Delivery teams were busy. Stakeholders were still showing up. Yet the conversations felt like defense, not discovery. Nobody could say, out loud, that the original rationale had changed. When a portfolio can’t stop, it stops being a portfolio. It becomes a backlog with a budget.
Stopping power is not about being harsh or impatient. It is about building an operating model where truth has a safe path to the surface, where commitments are reversible by design and where capacity can move toward real growth rather than protecting yesterday’s narrative.
MIT CISR describes an enterprise IT operating model as the accountabilities, processes, platforms, metrics and behaviors that define how the IT unit and business units collaborate. In the AI era, it must scale the use and reuse of data and AI while managing risks like cyber threats and privacy breaches. Stopping power is one of the operating-model behaviors many organizations are missing: The ability to reverse commitments when evidence changes.
Why portfolios lose stopping power
Stopping is difficult because the mechanics of enterprise work reward continuation.
First, money quickly turns into identity. After funding, leaders are no longer defending an idea; they are defending their credibility. The longer a program runs, the more it becomes a proxy for competence rather than just outcomes.
Second, governance often measures motion rather than learning. Status is tracked weekly. Evidence is rarely revisited. Early assumptions are treated as history, not as hypotheses that must keep earning their place. The enterprise excels at delivery updates but struggles with decision updates.
Third, decision rights are often unclear at the moment when matters stop. MIT CISR frames governance as the allocation of decision rights and accountabilities: Who has authority, who is accountable and how decisions stay coherent as you devolve power to teams. When those rights are fuzzy, stopping becomes a negotiation and negotiations default to “keep going.”
Finally, initiative volume keeps climbing. AI is making ideation cheaper. The portfolio inflates. Delivery capacity does not. Leaders end up managing overload through quiet deferral rather than explicit choices. That is how strategic work dies: Not through failure but through dilution.
Build stopping power into governance
Stopping power is not a speech. It is a system. The goal is to make stopping a normal outcome of good governance, not a scandal.
Require an exit plan before entry
If an initiative cannot explain how it will stop, it is not ready to start. I ask for four items upfront:
- A clear hypothesis: What must be true for this to be worth the investment
- A measurable time to value: What evidence we will see by a specific date
- A reversibility plan: How we unwind contracts, integrations, data and process changes
- A named decision owner: Who can say stop, and who must be consulted
McKinsey’s work on the evolving technology leadership mandate also emphasizes the shift from annual planning to more continuous decision making and tighter business-tech strategy cocreation. That shift only works if exit is treated as a first-class part of strategy, not an afterthought.
Install kill-switch gates throughout the delivery lifecycle.
The easiest time to stop is early, before a program acquires institutional gravity. I use three decision gates:
- Gate 1, viability: Is the problem still real, is the customer still there and is the baseline measured
- Gate 2, value signal: Is there evidence of adoption, cycle time improvement, risk reduction or revenue impact
- Gate 3, scale decision: Are the operating model, data controls, security and support model ready to expand
These are not stage gates that slow delivery. They are decision gates that prevent false certainty.
Separate portfolio truth from portfolio theatre.
Most steering committees mix three incompatible activities: Status reporting, issue escalation and investment decisions. That creates a bias toward performative confidence. A simple fix is to run a separate evidence review cadence focused on what changed:
- Which assumptions were tested
- Which signals improved or degraded
- Which constraints moved (regulatory, supply chain, talent, vendor terms)
- What does that imply for continuing, changing or stopping?
This is where stopping becomes normal, because the forum is built for learning.
Make capacity reallocation explicit.
Stopping without reinvestment looks like austerity. Stopping with reinvestment looks like leadership. Every stop decision should include a reallocation statement: What capacity is freed and where it is going. When teams see that stopping creates room for meaningful work, the politics shift.
Deloitte’s research on digital operating models highlights how reporting structure, ownership and team capabilities influence value realization. In practice, stopping power improves when ownership is clear, and the leader responsible for value has the authority to make hard calls.
Make stopping culturally safe in the AI era
Governance creates the mechanics. Culture determines whether people use them.
Use a narrative protocol that avoids blame loops.
If stopping is framed as “who messed up,” leaders will hide signals. If it is framed as “what did we learn and what is the best next bet,” leaders will surface reality earlier. The language matters:
- We are stopping because the hypothesis did not hold.
- We are stopping because the economy has changed.
- We are stopping because the dependency landscape shifted.
- We are stopping to protect focus and create capacity for higher value work.
Celebrate early stops as competence.
Many organizations reward persistence more than course correction. That is backwards. A well-run stop is proof of good judgment. A simple ritual helps: A quarterly “best stop” that highlights teams who proved something quickly, made a clear recommendation and freed capacity.
Protect teams from reputational shrapnel.
When programs stall, delivery teams often take the hit, even when the problem lies upstream: Unclear objectives, shifting priorities and missing decision rights. Leaders need to separate execution quality from investment quality explicitly. That is how you keep talent.
Now add AI, because it raises the stakes. BCG notes that GenAI can drive material productivity gains in the tech function and that tech leaders also need to help the broader organization scale responsibly while avoiding shadow IT. PwC’s 2026 CIO agenda likewise points to AI reshaping governance, operating models and culture. The practical implication is simple: Experimentation must be paired with clear kill criteria, guardrails and accountability. Without that, innovation becomes sprawl.
This is also where CEO alignment matters. The World Economic Forum has argued that CEOs set the vision for AI. At the same time, technology leaders build the processes that realize it, and that success depends on speed with structure, so momentum does not become chaos. Stopping power is the mechanism that turns that partnership into an execution advantage: It prevents the organization from scaling the wrong things.
A practical starter kit: The stop protocol pack
If you want to operationalize stopping power, start with a lightweight protocol pack and pilot it on a subset of initiatives next quarter.
Stop criteria
Stop when one of these is true:
- The business problem is no longer priority one.
- The hypothesis is not holding, and the next test is not worth the cost.
- Time-to-value is missed without a credible recovery plan.
- Risk or compliance exposure crosses a defined threshold.
- A dependency changes the economics (platform decision, vendor terms, data constraints)
- The program is consuming scarce capacity that could be used for a higher-value bet.
Kill switch gates
- Viability gate at 30 days
- Value signal gate at 60 to 90 days
- Scale gate before any enterprise rollout
Narrative protocol
- The decision owner communicates the stop with a one-page rationale.
- The delivery lead communicates what was learned and what assets are reusable.
- Sponsor communicates the reallocation plan and the next focus area.
- A short retro captures what to change in future intake, not who to blame.
Exit plan requirement
- Contract unwind path and timeline.
- Data and integration cleanup plan.
- User communication plan and transition support.
- Asset reuse plan (code, patterns, data models, controls).
If your next portfolio review has 30 initiatives and zero stop decisions, that is not a sign of excellence. It is a sign that truth has nowhere to go. Build stopping power, then use it.
This article is published as part of the Foundry Expert Contributor Network.
Want to join?