Peter Drucker’s famous observation that “culture eats strategy for breakfast” captured the power of organisational behaviour. Artificial intelligence may be preparing a breakfast for India. It is beginning to challenge the services engine that has underwritten India’s foreign exchange strength, while increasing dependence on imported technology and energy.

India’s latest export data presents an interesting contrast. Merchandise exports surged 23.04% to $44.40 billion in September 2026, while services exports were estimated at $36.35 billion, marginally lower than $36.67 billion a year earlier. Total exports of merchandise and services rose 10.99% to an estimated $80.75 billion. Services deserve attention because India has historically generated a dependable foreign exchange surplus here.

The monthly data shows the surplus declining from $18.3 billion in April 2026 to $16.53 billion in August, about 10%. If this continues, it raises a question about what happens to India’s external account when AI changes the economics of services exports.

Consider how the traditional services industry operated. It used accumulated knowledge of segments such as insurance or automobiles, deploying human labour with tools, techniques and processes to deliver services overseas.

In the standard theory of the firm construct, a firm can exist profitably when the cost of internal co-ordination to deliver an outcome is less than the cost of co-ordination through market mechanisms. The Indian services industry grew because it mastered internal co-ordination to cut costs. What is also striking is that many of India’s IT/ITES companies did not build significant R&D capabilities during their peak years, even as several of their founders’ family offices were investing in deeptech and emerging technologies elsewhere.

There is an import cost associated with this export. However, if exports increased by an additional 1%, the import cost did not increase to the same level, because local labour was deployed. This model had a desirable property: an increase in exports always resulted in an increase in surplus.

But what if AI disrupts this property? Firms are now delivering more with less human labour and more AI. Where this AI comes from, how it costs and in what currency matters. The answer falls into two buckets.

In the first bucket lies US inference providers such as OpenAI and Anthropic, whose agents are widely used in knowledge work, with costs measured in tokens and dollars. As Indian service firms replace more human labour with agentic labour, they import more dollars for every increase in dollar export. As a result, the surplus comes down.

In the second bucket lie Chinese open-weight models such as GLM and DeepSeek, deployed as is or with fine tuning on local data centres. However, India does not have a silicon manufacturing ecosystem for CPUs, GPUs and other components. These are imported and paid for in dollar. Running inference on locally hosted models at scale is harder, so firms usually gravitate towards the first.

Hence, irrespective of the bucket, firms can replace abundant carbon life-based labour with pricey silicon-based agents. As models become more capable, the disruption will accelerate and the dollar surplus from services exports will come down.

As the surplus earned by services exports decreases, it must be compensated for by surpluses from goods. The merchandise numbers provide encouragement, but the trade deficit remains large. Goods cannot compensate for a weakening services surplus.

This has a bearing on the exchange rate. The RBI typically manages the exchange rate by using its reserves. It is doing its role and has bought the Government of India time. But reserves are a stock, while external imbalances are a flow. India’s reserves have fallen sharply from their September peak. The pace of depletion matters.

The FCNR scheme has similarly bought borrowed time. The special swap window mobilised foreign currency and strengthened the RBI’s firepower. But these liabilities will eventually have to be returned. FCNR provides breathing space, not a durable source of foreign exchange. AI is now threatening the economics of human-intensive services globally.

The disruption also affects tax revenues. As firms replace internal co-ordination mastery with market mechanisms, their profitability dips because of increased costs and hence they pay less to the exchequer as taxes. Since fewer jobs are created, the multiplier effect reduces and leads to more demand for state support through subsidies and Direct Benefit Transfers. Political-economic considerations can then create fiscal deficits, adding another impediment to monetary action.

The RBI also has more than $100 billion mobilised via its FCNR scheme, with liabilities coming due over the next few years. It is also unclear how costly energy imports could become, given the Iran war and its offshoot conflicts such as the Houthi-Saudi war. Unlike China, however, India, still a lower middle income economy, has not embraced the transition to green energy and remains dependent on fossil fuels priced in dollar for its energy needs.

While the RBI will do what it can in this volatile poly-crisis era, it cannot manufacture foreign exchange. It can intervene, deploy reserves and use FCNR to buy time. India must use that time to build external strength before AI weakens the old one.

India’s services model was built by converting abundant human labour into dollars. AI is changing that equation by replacing labour input with imported silicon, compute and energy. Strategy may survive the breakfast table. Reserves may not.