Return on AI investment is a comparison of the benefits of a deployment with its total costs. Companies are often unable to quantify it for a simple reason: they did not measure the state before deployment, so they have nothing to compare against. The first step is therefore to record baseline figures – time to process a task, number of errors, volume of requests handled, cost of external supplies. On the cost side, do not forget the ones that are not stated in offers: data preparation and cleaning, integration with existing systems, running costs, the time of internal people, training, oversight of outputs, and ongoing maintenance when models change. Calculate the benefit conservatively, and do not confuse time saved with money saved – savings translate into profit only if the freed-up capacity is actually put to use elsewhere.
See also: Testing before investment, Cloud services in a project, KPIs for a marketer.
