Research commentary by Crestlane. The figures describe a global survey sample, not the Indian market or Crestlane client results.
Source: Stanford 2026 AI Index, chapter 4.3, reporting McKinsey survey data. Self-reported adoption; not an India-specific estimate.
What the research measures
Stanford’s 2026 AI Index reports that 88% of respondents in the underlying organisational survey said their organisations used AI in at least one business function in 2025. The comparable figures were 78% in 2024 and 55% in 2023. The report draws on McKinsey’s annual survey for this series.
These are self-reported adoption figures. They do not tell us how many workflows operate without manual intervention, whether a deployment is profitable, or how adoption differs in an individual Indian industry. One function using AI is a much lower threshold than an entire business operating differently.
Our reading: the next question is operational
For a business considering an investment, “Are we using AI?” is a weak planning question. A more useful one is: “Which recurring decision or handoff are we trying to improve?” It points the discussion toward a process, an owner and a measurable baseline.
Consider a sales enquiry. A model can classify the message, but the workflow still needs to match an existing contact, decide where to route it, preserve the original message and tell someone when the integration fails. The model is one component of the system. The surrounding decisions determine whether the sales team can depend on it.
Build a baseline before a pilot
Choose a process with enough repetition to study. Record its incoming volume, the time spent handling a typical item, how often someone has to correct it, and where work waits between people. Measure the same things during the pilot. Avoid combining a busy week and a quiet week into a single dramatic percentage.
Agree which outcomes require review. A suggested email can wait for a person to approve it. A record update may need a validation rule. A payment instruction needs a different level of control. Treating these actions as equivalent makes a pilot hard to evaluate and harder to operate.
A practical investment decision
We recommend a small decision document at the end of a pilot: what changed, what still fails, what it costs to run, and who will maintain it. Include the time spent reviewing and correcting the output. A useful system should improve the total workflow, including that human work.
The adoption chart is context for the discussion. Your own process evidence should determine the investment. Start with one repeatable workflow, define a review boundary, and expand only when the people operating it can explain how it behaves.