For decades, long-term investing has been evaluated using the same toolbox: NPV, IRR, best/base/worst case scenarios, and sometimes a little real options to make it look serious. A useful grammar – but based on a comfortable lie: the world is stable enough for our assumptions to remain so.
The academic article Artificial Intelligence in Corporate Financial Strategy: Transforming Long-Term Investment and Capital Budgeting Decisions
Authors by Ashutosh Roy et al. (JEFAS, 2025) makes a simple point: AI doesn’t replace corporate finance – it makes it practicable in an unstable world. And that changes everything, because in the real world, investment decisions aren’t problems of calculation: they’re problems of revisibility.
What literature really says (and implies)
1) Classic methods are not “wrong” – they are rigid
NPV and IRR remain fundamental because they impose a discipline: compare, update, decide. But they have inherited structural constraints:
- fixed assumptions (rates, cash flows, competitive stability)
- dependence on history (which is precisely what is least reliable at break-up)
- difficulty integrating weak signals (regulation, sentiment, techno, ESG)
In other words, they’re perfect… for a slow-moving world.
2) AI is not a “better Excel”, it’s a regime change
Roy et al. insist on a breakthrough: AI enables dynamic models (continuous updating, external signals, non-linearities). It intervenes where conventional finance suffers:
- cash flow forecasting via ML / time series
- large-scale scenario analysis (massive simulations rather than 3 hypotheses)
- integration of non-financial data (NLP on news, ESG reports, sentiment)
What this produces is not just “more precise”: it’s more adaptable.
The point that really interests a COMEX: AI restores flexibility… and values it
Finance has always had this idea (real options): the real asset is the ability to defer, extend, reduce or abandon. Problem: it’s expensive to model, and therefore rarely done properly.
The article makes a key point: AI makes this flexibility operational in real time. An option is no longer a theoretical chapter; it becomes a signal-driven mechanism:
- market volatility
- competitive pressure
- technological disruption
- supply / energy constraints
- regulation and ESG
As a result, we no longer just value a project, we value a decision trajectory.
The Tesla case: useful… but to be read as a demonstration, not as proof
The authors use Tesla (FCF over 5 years) to illustrate the difference between traditional and AI approaches.
- Classic” NPV: +$7.56 bn (fixed assumptions)
- NPV “IA”: +$8.95 bn (growth adjustments + efficiency gains)
- probability of positive NPV: 72% → 88% via simulations
It’s instructive: you can see how an “AI forecasting + Monte Carlo + ESG signals” layer shifts the distribution of results.
But beware of reading bias: it’s not “AI magically discovers $1.4 billion”. It’s more like:
AI formalizes what decision-makers already do intuitively (revise assumptions, integrate external signals, reweight risk) – and makes it calculable, traceable and auditable.
Where the article becomes (unintentionally) explosive: governance, responsibility, explicability
The ethics/governance section is the real nerve of the war, because it points out the paradox:
- AI increases analysis power
- but it can reduce the capacity for explanation (“black box”)
- and amplify historical biases (imperfect data, incentives, ESG greenwashing)
In an investment decision, this is not a detail. A decision that can’t be explained is a decision that can’t be defended – and therefore can’t be governed.
The literature converges on a few answers:
- Explainable AI (XAI ): making variable contributions intelligible
- model audit: drift, bias, robustness, stress tests
- board-level monitoring: who signs? who reviews? on which triggers?
- data governance: provenance, quality, compliance, use of “alternative” data
The subtext is clear: AI is not an analyst’s tool. It’s a governance issue.
What marketers need to capture (or miss)
Because yes: “corporate finance” sounds a long way from marketing. In fact, it goes right to the heart of marketing investment allocation.
AI applied to capital budgeting sounds exactly like what marketing teams are experiencing:
- uncertain cash flows (incremental revenues, CLV, churn)
- unstable assumptions (CAC, saturation, media mix, competition)
- a need for revision (stop rules, reallocation, timing)
The actionable lesson is as follows:
The future is not “AI does budgeting”. The future is “budgeting becomes revisable by design”.
And that implies refereeing discipline:
- Explicit assumptions (what must be true for the investment to hold)
- Massive scenarios (not 3 variants – one distribution)
- Managerial options (what we can change if the world moves)
- Reopening conditions (dated, measurable, enforceable triggers)
- Explicability (to defend, learn and iterate)
Conclusion: AI doesn’t “decide” better. It makes the decision reviewable without losing face.
Roy et al. defend a strong thesis: AI is becoming a transformative force in financial strategy, as it improves forecasts, simulations and ESG integration, while raising issues of transparency and governance.
