Portfolio leaders may be misreading signals. For example, a portfolio team reviews ten AI investments from last year. Six produced no measurable return. The analysis shows a 60% failure rate, and the recommendation is to pull back.
That conclusion may be wrong, and acting on it might be expensive.
Luke cautions leaders to avoid acting on data analyzed in isolation which cannot indicate if the portfolio is working. What matters is how a well-built portfolio is supposed to behave in the first place.
AI has not changed the fundamentals of portfolio management. It has made the stakes higher, the cycles shorter, and the mistakes costlier.
Luke's view of innovation investing draws on three roles he has held: Silicon Valley entrepreneur, angel investor, and limited partner in a venture capital fund. Each one teaches the same lesson about innovation bets.
The venture rule of thumb is simple. Out of 10 investments, roughly six fail, two or three return a modest amount, and one becomes the unicorn that pays for everything else. Angel investors fail even more often; their wins are also much larger.
So a 60% failure rate in Horizon Three is not evidence of waste. It is the expected shape of a portfolio that is pushing the envelope.
The failures still pay something back. If only two or three of 10 initiatives earn a move into Horizon Two, the rest produce learning: a clear signal that the idea is not a candidate for productization. That signal has value, and it arrives sooner and cheaper than a failed product launch.
The gate into Horizon Two is traction. In Luke's words, VCs invest in success. A seed-stage bet becomes an A or B round only when it shows growth; a Horizon Three initiative should move to Horizon Two on the same evidence. That also means Horizon Two should stay small.
One more implication: the definition of ROI shifts as an initiative matures. Early on, the return is validated learning. Later, it is revenue, margin, and share. A portfolio that applies Horizon One expectations to Horizon Three bets will kill its future before it has a chance to grow.
Horizon Three gets the headlines. Horizon One pays the bills.
The products already making money need ongoing investment, and that investment should be funded by their own revenue. Right now, much of that money is going into AI. Existing products are being overhauled to become AI-native, not just patched with a new feature.
Applied Frameworks is living this decision. Horizon Engage started as Weave, was sold to Scaled Agile, where it became SAFe® Collaborate, and was bought back last year. The team is now investing new AI capabilities into a product with a long track record.
This creates one of the hardest questions portfolio leaders face today. Should funding go to a new Horizon Three bet, or to making a Horizon One product AI-native? Both are reasonable. Some of each will work; some will not.
The only way to compare them is to model both through the same ROI lens. Every investment sits on a J-curve: it dips before it returns, and some curves pay off in months while others take years. Horizon Invest makes that curve visible across Core, Growth, and Transform, so the comparison is based on cost, timing, and return rather than whoever argues hardest.
The pressure is real. There has always been more opportunity than resources to pursue it. AI has increased the size of the opportunity bucket; the resource bucket may not have grown to match. That could rule out the default of last year's budget plus or minus 10%. Every investment has to earn its place again each period.
Annual planning cycles that take six months to complete are not a competitive advantage. The most progressive portfolio leaders are moving to quarterly investment reviews, with room to adjust funding mid-cycle when conditions change.
Quarterly is not arbitrary. ARTs may be moving faster than ever, but the financial world still reports on quarterly earnings, and external market forces set a rhythm of their own.
The timing detail matters. In the LPM cadence, which carries into AI-Native SAFe, portfolio decisions happen on a quarterly cadence that is deliberately offset from the ART cadence. The portfolio decides first; the ARTs implement on their next boundary. That gap gives ARTs time to prepare, and it gives HR and other functions time to prepare too.
Faster cadence also opens the door to a bigger change. Over time, ARTs get messy: a solution splits into two market segments, or a new product spins off. Adopting AI-Native SAFe is a chance to reshape ARTs so each one is sized for the solution it builds. Sometimes that means a larger ART and a restructured solution train. Sometimes it means several smaller ARTs, and a new funding question at the portfolio level.
None of this works with spreadsheets and slide decks. A faster cadence requires a different set of tools, which is exactly what Horizon Invest is built for: quarterly or on-demand reviews, with allocations that move as business conditions evolve.
The industry talks about keeping humans in the loop. Luke goes a step further: leaders in the loop, or leaders above the loop.
Agents can help shape strategy. They can track progress. They cannot be fiduciaries. There is no legal framework in which an agent makes a multi-million-dollar investment decision and the corporation simply accepts responsibility for it. When the decisions often run into tens or hundreds of millions of dollars, humans remain accountable, and the board holds that accountability.
So the question is not whether AI makes the decision. The question is what leaders need to make better decisions, faster. Luke names three needs:
The second need is where most portfolios fall short. The people closest to customers and to the work often have the insight leaders need, but there has never been a scalable way to collect it.
That does not mean every decision becomes a vote. A merger, acquisition, or divestiture has to stay confidential; for a public company, a leak can move stock prices and violate SEC rules. But a decision about new AI capabilities in a Horizon One product could benefit from input across the entire development team.
Luke has seen this pay off. At a previous company, three of the five best features in the platform came from the development team, after leadership took the time to share what customers needed, where the company was headed, and what it would not do. AI raises the stakes on that practice. The same intent, context, and guardrails now go into the specifications agents work from, so people and agents operate from common information.
Horizon Invest is designed around this idea. Decision sessions are structured and auditable, with assumptions, trade-offs, and adjustments captured in real time. Participants see the same investment options, constraints, and priorities. Product, finance, engineering, and executive leaders work through trade-offs together, with individual votes followed by group consensus.
Proving ROI starts before the money is spent. It starts with a business case that states clear hypotheses.
In a traditional SAFe portfolio, building a strong business case could take weeks. Working with clients on the Horizon platform, Applied Frameworks has cut that to one or two days, using agents to help model the investment. The output is not a single-number guess. It is a set of explicit hypotheses and a thorough, range-based Monte Carlo analysis.
That speed changes what a portfolio can do. Once the hypotheses are written down, each one can be proven or disproven. Each proven or disproven hypothesis moves the organization closer to proving, or disproving, the ROI itself. Then the portfolio can ask the next question: even if this investment is delivering, how does it compare with the other opportunities now in view?
There is one more test that most organizations skip. At the Applied Frameworks booth at the SAFe Summit, Luke asked visitors a simple question: when was the last feature you delivered that was so good you were able to raise your prices?
Most people could not point to a specific example. They had been shipping value steadily without revisiting pricing, packaging, or licensing. Customer centricity is the foundation of SAFe, and it should be. But a sustainable business also has to monetize the value it creates. The best price increases are driven by value, not by inflation or competitors.
This is the gap Horizon closes. Built to include the Profit Streams framework, it connects funding decisions to delivery progress and to the profit each initiative actually produces. Portfolio scoring uses realized profit, not guesswork.
Luke often goes back to Peter Drucker: the purpose of a company is to serve its customers. That has not changed. What has changed is the speed of the loop.
Understand the customer. Track how they are evolving. Bring that insight into decisions. Know which horizon each investment belongs to. Review the portfolio on a cadence that matches the market. Capture the value shipped as profit. Then do it again, faster.
AI accelerates every step. It does not remove the need for leaders to own the outcome.
If you are heading into budget season and want to see how Horizon Invest supports quarterly, auditable portfolio decisions, book time with me and let's discuss!