ADLC: Breaking the Portfolio Bottleneck

I was speaking with Shannon yesterday about SPM and ADLC, and our conversation encouraged me to publish this post a little early. So, Shannon, this one is for you!

Strategic portfolio management is the last function still planning on an annual clock. The tooling is about to make that indefensible.

Part of the ADLC series. “From SDLC to ADLC” was how the work changes once AI is in the build. This is where that work comes from.

BLUF: Nearly every part of the enterprise has embraced some sort of continuous execution and delivery except the portfolio. From a technical execution standpoint, delivery ships multiple times per day, models retrain on production telemetry, and security detects and responds to threats in near real time. And the portfolio that governs all of it still plans on an annual budgetary cycle, defends business cases at a gate, and scores success as conformance to a plan written before anyone knew what they would learn. We tried to address some of this with Lean Portfolio Management, but even in the best adoptions, the system iterated only two to four times per year. Truthfully, in 2026, that is too slow. The next generation of SPM tooling aims to close that gap, not by helping us build the annual plan faster, but by making the annual plan a thing of the past. What follows is a mix of what the vendors are actively doing and what they should be working on in their respective labs. The shift is from monitor-and-control to sense-and-respond, and the organizations that make it will run their portfolios at the speed their business already moves. This is the strategy layer of that lifecycle: the portfolio decides what is worth building, the ADLC decides how to build and run it, and right now the two keep different time.

I have sat in the annual planning room both as an advisor and as an executive with P&L ownership. Two days offsite, a wall of initiatives, a spreadsheet that took a quarter to assemble, and a set of business cases each defended with unshakable confidence. By the time the binder is approved, at the pace of business today, the assumptions underneath half of it have already moved. By Q2 the plan is a historical document. By Q3 everyone is working off a different list and pretending the plan still governs. (I have helped build a few of those perfect plans. I am not proud of all of them.)

Let’s be clear: this is not a planning-quality problem. You won’t simply “plan better,” and your enterprise or industry is not uniquely insulated from the reality of the modern pace of business. The problem is that the portfolio is being operated like a road trip. Set a target, measure variance, correct back to the plan. That works when the plan is a good description of reality for twelve months, but it stops working the moment the thing you are governing changes faster than once a year.

And everything the portfolio governs now changes much faster than once a year.

The portfolio is the front door to the lifecycle

I have written about the shift from SDLC to ADLC: how building and operating AI systems runs on different physics than traditional software, where production is a feedback source, there is no “done,” and the work is measured by learning velocity instead of predictability. The portfolio sits above all of that. Large bodies of novel work and new experiments have to enter the lifecycle from somewhere, and increasingly that somewhere is the last place to deploy agents. That has to change.

In an agentic delivery model, a growing share of decisions gets made close to the work. Coding agents refactor, test, and ship inside guardrails a human set. Teams run experiments daily. That is the point of the ADLC: push the decision to where the context lives and let the work move at the speed the tools now allow.

But not every decision belongs close to the work. New bets do not. Significant refactors do not. Infrastructure that changes the cost or risk profile of everything downstream does not. We refer to this work as having significant economies of scale, and these bodies of work are still portfolio considerations, because they are choices about where to point the money and the attention before anyone, human or agent, starts building. The portfolio is, and will continue to be, the bridge between strategy and execution. It decides what is worth doing and funds it against strategic intent. The ADLC decides how to do it, and does most of that deciding itself.

So, the portfolio is the front door to the lifecycle. It is where strategy enters and becomes work. And if the lifecycle downstream of it now runs every day, a front door that opens once a year is the constraint on the whole system.

The business case stops being a one-time artifact

Starting with a lightweight business case is a valuable thought experiment, because these initiatives are the very thing that can break the annual model most visibly, and where the tools can help first.

Today a business case is a document that most will write once, defend at a gate, and then spend the next twelve months explaining variances against. It is an argument made at the moment of maximum ignorance: before the work starts, before the first experiment, before anyone has touched the data. We have built an entire governance apparatus around treating that first guess as the baseline of record.

The next generation of SPM platforms changes the input side of this. Given a clear statement of strategic intent, a set of budgetary guardrails, and enough learned context about how value moves through the business, where work piles up, and how the architecture is wired, the platform can surface the work that advances the intent. The candidate work itself, derived from what the tools know of your strategic intent, the monitors established for the market, and how value flows and where it gets stuck, proposed as a set of experiments with a hypothesis attached to each one.

That is a real change in what a business case is. It stops being a defense of a predetermined answer and becomes a continuously re-derived view of the highest-value experiments available right now, given what the organization has learned since the last time it looked.

I want to be clear about the dependency here, because this is where the vendor demos can get ahead of reality. A platform can only surface the right work if it can see value flow, bottlenecks, and architecture clearly, and most enterprises cannot. The value-stream data is fragmented across tools and the architecture-to-value mapping, in most cases, does not exist. Is the platform that reads all of that and proposes the right experiments shipping today? Not yet. This is the part still in the labs. The foundation is there and the direction is set, but the capability arrives on top of instrumentation most organizations have not yet built. The firms that get there first will be the ones that treated value-flow visibility as infrastructure, the same way the firms getting real value from AI treated production observability as core development work rather than an operations afterthought.

This is also why the platforms best positioned for it are the ones already sitting on the operational system of record. ServiceNow SPM is the clearest case: it runs on the same platform as the ITSM and ITOM data, so the work and the application landscape already live where the portfolio governs, and its Digital Portfolio Management ties applications back to business capabilities. Proximity to the operational data is what makes surfacing real candidate work plausible instead of aspirational. The platform that can see where the work moves has a head start over the one staring at a tidy list of approved projects.

Iterate at the speed of the business, not the speed of the calendar

Once the business case is something the platform helps re-derive continuously, the annual cycle has very little left to do.

Let me be exceedingly clear on one point so this does not get misread. We may still set an annual capital envelope, and we will still set annual sales targets. That part is fine and it is not going anywhere, because fiduciary oversight and forty-year-old capital budgeting frameworks run on a fiscal calendar, and strategic intent is framed on a longer horizon anyway. What has to end is predefining the entirety of that budget up front, committing every dollar to a named initiative twelve months out, which is exactly what most organizations still do. The world is changing too fast for it. I cannot tell you what the priorities should be ninety days from now, let alone a year out. Nobody can. So, you set the guardrails and the intent, then point the money at the evidence as the year unfolds. The guardrails stay; the frozen plan inside them goes.

What changes is everything inside those guardrails. SAFe’s Lean Portfolio Management, a body of knowledge that I am proud to have contributed to, pointed the right way as early as 2019: fund value streams to guardrails rather than projects to detailed business cases, re-sequence with Weighted Shortest Job First as you learn, and reallocate at cadence-based portfolio syncs instead of waiting on a steering committee. It worked, but it has a ceiling. Even strong adoptions of LPM re-plan their portfolio only two to four times a year. Though this was a generational improvement on annual planning, in 2026 it is still a batch process. The next generation of tooling drops the batch: reallocation becomes event-driven, triggered by evidence instead of by the calendar or the cadence.

The result is a portfolio that re-sequences itself as evidence arrives. An experiment validates in March, the funding follows it in March. An assumption collapses in June, the money moves off it in June. The portfolio iterates at the speed the business is already moving, instead of holding what it learns in a queue until the next planning season opens.

Moving money in March instead of next January sounds reckless right up until you can see the second-order effects before you commit. That is what scenario modeling is for, and it is where the financially serious platforms have spent their effort. Planisware Horizon is the one I reach for here, built around what-if modeling across funding, capacity, and risk: you simulate the reallocation, watch what it does to the rest of the portfolio, then decide. The skill that used to live in one finance analyst’s spreadsheet becomes a standing capability of the portfolio. Continuous reallocation is only responsible when you can rehearse it first.

From monitor-and-control to sense-and-respond

This is the operating-model shift underneath all of it, and it deserves its own name. Portfolio operating is moving from monitor-and-control to sense-and-respond.

Think about how air traffic control works. Controllers do not write a plan in January that routes every flight for the year and then measure variance against it. They set a destination and a safety envelope, then they sense conditions continuously and re-route in real time within that envelope. The guardrails are strict and non-negotiable. The paths inside them are adjusted constantly. Nobody calls a re-route a failure of the plan. Re-routing is the plan.

A monitor-and-control portfolio treats every deviation from the original plan as a variance to be explained and corrected. A sense-and-respond portfolio treats deviation as signal. The strategic intent is the destination. The budgetary guardrails are the safety envelope. Everything in between is expected to move as conditions change, and the job of governance is not to prevent movement but to keep it inside the envelope. Those guardrails do double duty. They are also the line between what the work decides for itself and what comes back to the portfolio: inside the envelope the teams and the agents move freely, and the moment a choice would breach it, a new bet, a major refactor, a shift in the risk profile, that choice belongs at the portfolio.

Sensing requires something to sense. A portfolio cannot respond to a signal it cannot see, which is why the platforms are racing to wire the portfolio layer to live delivery data. Broadcom’s Clarity is the clearest version of that bet: paired with Rally under the ValueOps banner, it is built to tie portfolio decisions to real-time execution telemetry instead of to a status slide assembled the night before the review. The detail I find telling is the marketing. Broadcom now tells buyers to stop wasting months on rigid annual plans and steer on a quarterly cadence instead. Quarterly is still a batch, and still slower than where this ends up, but the tell is the direction of travel. When a forty-year-old PPM incumbent is arguing against the annual cycle, the question stops being whether the model changes. It becomes who restructures around it first.

This is the same physics I wrote about in the shift from SDLC to ADLC, moved up a level. In the AI Development Lifecycle, you stop asking “did we finish the build phase” and start asking “did we validate the business hypothesis that justified this investment.” The portfolio version asks the same question across every value stream at once: what are we learning, and what is it telling us to fund more of and stop funding. Those are the right questions. Most portfolio reviews are still asking whether we are on schedule and on budget, which are the right questions for the wrong problem.

One portfolio, governed by value

The last shift is the one most organizations will resist hardest, because it dismantles a boundary that has org charts and career tracks built on top of it. The technology portfolio and the operations portfolio become one portfolio, organized around value delivery rather than around who owns the work.

The separation made sense when building software and running it were different activities done by different people on different clocks. That separation is the same one ADLC dissolves at the team level, where the production signal becomes a development input and “build and done” becomes “build and run.” At the portfolio level the same logic holds. If the model in production is still a development artifact, then operating it is not a separate cost center to be managed against an SLA. It is part of the same value stream, funded continuously, governed by the same outcome-based guardrails.

Here is where I will add a caution the enthusiasts skip. Merging the portfolio view is right. Merging the guardrails wholesale is not. Run-the-business operations fail in continuous, sometimes safety-critical ways that change work does not. A regulated operations failure is a different risk class than a delayed feature. The defensible model is one portfolio and one value lens, with differentiated guardrails for the run-the-business work and the change work underneath it. Same board. Same prioritization logic. Different safety envelopes. Collapse that distinction and you will learn why it existed, probably from your regulator.

What this asks of you

None of this is a tooling purchase. All three are solid. They come at the problem from different corners: ServiceNow from the operational system of record, Planisware from financial and scenario planning, Clarity from the value stream. Pick the one that fits your stack. None of them hands you the operating model. The platforms make the continuous portfolio possible. They do not make it happen, any more than buying a flight-tracking system makes you an air traffic controller. The work is structural, and it is the unglamorous kind.

Fund value streams to guardrails instead of projects to business cases. Move reallocation decisions off the calendar and onto evidence. Rebuild governance around outcome validation instead of phase completion. Instrument value flow so the platform has something real to learn from. Put the run and the change work on the same board with different safety envelopes. Every one of those is a conversation with someone whose job was built around the model you are changing, which is why most organizations announce the tool and skip the work.

And there is a cost to skipping it that never shows up on a portfolio dashboard. It is the team that learned something real in Q2 and had to sit on it until next year’s planning window opened, watching the opportunity close while they waited for permission to act on what they already knew. It is the leader who stops bringing new ideas to the portfolio because she has learned the cycle cannot absorb them between planning seasons. It is the strategy that was exactly right in January and irrelevant by June, executed faithfully the whole way down because the portfolio had no mechanism to notice it had stopped being true. Annual planning does not just slow you down. It quietly teaches your best people to stop surfacing what they learn.

The portfolio has been the slowest-moving part of the enterprise for a long time, and for a long time that was tolerable, because everything it governed moved slowly too. That era is over. The ADLC made delivery continuous and pushed a generation of decisions down to the work. What it did not change is where the big bets come from. That is still the portfolio, and a front door that opens once a year cannot feed a lifecycle that runs every day. Strategy and execution have to keep the same time now. The only question left is whether the operating model catches up, or whether the portfolio stays a binder that everyone has quietly stopped believing by spring.