No business ever learns what it is actually good at investing in
Module · Capital and Investment
Every investment is approved on a promise and almost none is checked against it, so the same optimism is priced into the next paper and the organisation never discovers what it is good at.
Prophesee Corporate keeps the case live, predicts what this business will actually deliver rather than what the paper claims, and turns a history of decisions into an advantage in the next one.
Half of large projects land on budget. One in two hundred lands on everything.
Across the largest database of major projects assembled, just under half come in on budget or better, and only half a percent come in on budget, on time and with the benefits that were promised. The approval process is rigorous. What is missing is the loop back, so the same optimism is priced into the next paper.
Sources: Bent Flyvbjerg and Dan Gardner, How Big Things Get Done, 2023, drawing on a database of more than 16,000 large projects, building on Flyvbjerg, What You Should Know About Megaprojects and Why, Project Management Journal, 2014 · Project Management Institute, Pulse of the Profession 2025, published 2025 (n=2,841 project professionals worldwide, fieldwork Jul to Sep 2024) · 3RDi review of published capital budgeting and post-investment review research, Aug 2026, which found no sampled study of post-investment review frequency.
From approving the case to keeping it alive
ForesightEvery case is built once, by the team that wants the money, on assumptions nobody scores afterwards. The optimism is invisible because it has never been measured.
Models trained on every investment this organisation has completed, capital, technology, acquisition or programme, predict what an approval will actually deliver given its type, sponsor, size and assumptions, with a confidence band. The paper is judged against what this business has achieved, not against what it hopes.
PulseA project is reviewed at the gates. Between them the assumptions that justified it can move a long way without anyone being told.
The assumptions in the approved case become watched objects: volume, price, timing, adoption, unit cost. When an assumption moves outside the range the case depended on, the alert is raised against the case rather than against the project plan, and it is routed to the sponsor.
HorizonProjects are ranked by a return each sponsor calculated for themselves. There is no view of the portfolio, and no no-action baseline to compare any of them against.
Model the portfolio under constraint: capital available, delivery capacity, sequencing, risk appetite. Watch the projected return update as investments are added and removed, then track actuals against the funded curve with counterfactual modelling separating the investment from the market.
NexusCases live as documents in a folder. Nobody can query what was promised, who approved it, on what assumption, or how similar cases turned out.
Business cases, approval minutes, contracts and actuals resolved into one graph, with assumptions extracted from the documents by language models. Ask what was promised, what was delivered and what the difference was, for any investment or any sponsor.
Turning capital challenges into decisions
Projects ranked by returns each sponsor calculated for themselves, with no portfolio view and no no-action baseline.
HorizonBuilt once by the team that wants the money, on assumptions nobody ever scores against outcome.
ForesightCapital is allocated by division because that is how the budget is structured, not by where the next pound earns most.
ForesightCases live as documents in a folder. Nobody can query what was promised, by whom, on what assumption.
NexusHeld rarely, and when it is held the shortfall is attributed to the market, because nothing can separate the two.
Foresight Horizon13 AI applications that could be relevant
A sample of what becomes possible on the decision layer, not a fixed list: each application draws on the same data foundation and audit trail, and new ones are configured on the engines, not built from scratch.
The return this organisation will actually achieve, predicted from its own completed investments.
How far a case sits from what this business has actually delivered on investments like it.
Where the next unit of capital earns most, by asset, division and initiative.
The counterfactual separating what the investment delivered from what would have happened anyway.
Return on capital employed at the level decisions are actually made.
The assumptions the case depended on, watched, with the alert raised against the case.
Capital drawn against capital approved, tested continuously by project and sponsor.
Model the portfolio under capital, capacity and sequencing constraints and watch the return update.
The approved case tracked against actuals all the way to the delivered return.
Test the order in which investments are started, not just which ones are chosen.
Every case, assumption, approver and outcome, queryable in plain language.
What was promised against what arrived, for any investment or any sponsor.
Which assumption broke, by how much, and whether it breaks in every case of this type.
A day when the committee knows what it delivered
Today: A business case built by the team that wants the money, scored by nobody afterwards.
ForesightThe case in front of her compared with what comparable investments here have returned, with the gap named.
Today: An assumption that moved in month four, discussed at the gate in month nine.
PulseAdoption is running below the range the case depended on. The alert is raised against the case, not the project plan, and routed to the sponsor.
Today: Rank the projects by returns each sponsor calculated for themselves.
HorizonTwo investments swapped under the same capital and delivery constraint. The projected return moves. So does the risk.
Today: The programme finished. Whether the benefit arrived is a matter of opinion.
HorizonLast year's programme tracked to delivered return, with the counterfactual separating it from the market. That goes to the board.
Tomorrow's return becomes today's decision.
Fund what actually works
We agree the metric and the baseline in week one, and measure the result on your data.