ArthaVeda view for business owners to assess favourable odds on pitched AI opportunities.
If you own a business, you have probably been pitched on AI in the last six months. A banker brings a transaction that "positions you for the AI era." A consultant offers a transformation programme. A vendor shows a demo that makes your operations look slow and old. Everyone agrees you must act, and nobody mentions that acting and winning are different things.
This post gives you one way to think about it. Not every business needs AI, and not every business that spends on AI will earn anything from it. The question ArthaVeda asks before investing is not "how much should we spend?" or even "should we spend?" The first question is: does this investment have favourable odds? If it does, you are investing. If it does not, you are speculating, however sophisticated the pitch sounds.
Why the odds come first
Most AI proposals arrive with a budget, a timeline and a projected return. The return is usually built on assumptions about savings or growth that no one has tested. Discounting those cash flows at your cost of capital makes the spreadsheet look rigorous, but it does not make the cash flows real.
That is why the order matters. First decide whether the opportunity has favourable odds. Only then ask whether the expected payoff clears your hurdle rate after counting ongoing costs such as compute, talent and model upgrades. If the odds fail, there is no point in modelling the returns, because the cash flows you would be forecasting are not reliable.
Favourable odds depend on three factors: durability, optionality and a catalyst. Any one of them, genuinely present, is enough to make the odds favourable. If none is present, the spend is a gamble. All three factors are grounded in earnings.
Test 1: Earnings durability
Will AI make your earnings harder for competitors to attack, in a way they cannot easily copy?
AI models and tools are available to almost everyone. If your plan is to buy the same tools your competitors can buy next quarter, you have bought access, not advantage. Whatever gains the tools produce will usually be competed away through lower prices, with the benefit passing to customers instead of showing up in your margins.
Durability tends to come from pairing AI with something you already own that rivals lack:
Proprietary data that improves with use, such as years of transaction records, equipment readings or customer behaviour that no one else has
Distribution or workflow lock-in, where AI sits inside the way your customers already work and is costly for them to leave
Scale, where the fixed cost of the AI system is spread over a much larger base than a smaller rival could manage
A simple check: if a competitor could replicate the result within a year with the same off-the-shelf tools, the earnings are not durable.
Test 2: Optionality
Does one spend open several earnings streams?
The strongest case for AI investment is reuse. A well-built data foundation, a trained team and a deployed platform can serve many purposes. The first use case carries the cost, and the second and third come cheaply. One investment, many possible earnings streams.
To check for optionality:
Name at least three distinct earning streams, such as lower costs, a new product, better pricing or improved customer retention, that run on the same investment
Separate the shared spend, which every use case can draw on, from the spend that serves only the first use case
Ask what portion of the budget buys something reusable and what portion buys a one-off tool
If the proposal is a single-purpose tool with no reuse, it has no optionality. That alone does not make it a speculation, but the spend must then pass on durability or a catalyst.
Test 3: A catalyst
Is there a specific trigger that will lift your earnings, and can you say when?
"Productivity gains" and "operational efficiency" are not catalysts. They are hopes. A catalyst is a specific mechanism that moves a specific line of your profit and loss account, such as a pricing change, a product launch, a measurable fall in cost per unit, or a clear improvement in customer retention.
Ask three things:
Which line of the P&L moves, by how much, and by when?
What leading indicator will tell us within a few quarters whether it is working?
Does the catalyst depend on customers, regulators or partners acting in a certain way? If so, the uncertainty is higher.
If no one can answer these questions, the spend has no trigger. It may produce activity, but there is no reason to expect it to produce earnings.
At least one must be present
One real factor is the minimum for favourable odds. You do not need all three, but each one you add strengthens the case, and each one that is missing leaves a specific weakness you should price in:
Optionality without durability still gives you several earnings streams from one spend, but rivals can build the same streams on the same tools, so each one is easier to compete away.
A catalyst without durability still gives you a specific lift in earnings, but rivals can copy it, so expect the lift to be temporary.
Durability without a catalyst still protects the earnings you have, but nothing in the plan lifts them, so the gain lies mostly in not losing ground.
Durability and a catalyst without optionality is a strong case, with the lift defended, but it rests on a single stream, so the upside is limited to that one use.
When at least one is present, and you can show it with evidence, the odds favour you. You are investing, and you can move on to valuation and hurdle rates. The more factors present, the stronger the case. When none is present, you are speculating on AI.
A small bet can still be speculation
A tempting response to uncertainty is to start small: run a pilot, phase the spend, keep it cheap. These are sensible risk controls, but they do not change the nature of the bet.
Speculation is not defined by the size of the stake. It is defined by the structure of the odds. A small bet on a proposition with no durability, no optionality and no catalyst is still speculation. It just loses less when it fails. Phasing limits how much you can lose, but it does nothing to improve the probability that you win.
There is one legitimate use of a pilot: to test whether the three conditions are true. For example, you might test whether your proprietary data improves model results. But then you are spending for research into whether the odds are favourable. The full commitment should wait for the answer. Do not let a pilot become the first instalment of a spend you have already decided to make.
The bottom line
Investment on AI is not inherently wise or foolish. What matters is whether the spend has favourable odds, and that comes down to three things: whether it makes your earnings more durable, whether it opens more than one stream of earnings, and whether it sets off a specific lift in profit. Any one of them is enough for favourable odds.
A big payoff does not change this. When a banker or advisor shows you a large return or a huge market opportunity, that tells you what you could win, not how likely you are to win it. Without favourable odds, that directly improves earnings, a big payoff is a lottery ticket: the prize is attractive, but the odds are against you. Ask about the odds before you look at the prize.
Capital Decision are hard to undo. Whether you are looking to raise money for AI investment, or evaluating a proposal to invest in AI, we advise you on the full picture. Before you commit, talk to Trishula Capital by filling the Contact Us form.
