The Prediction Is the Easy Part
An interview with Eva Narunovska by Sean Murphy
Sophisticated forecasting has always been something you bought with scale. Predicting which customers will leave, which deals will close and what to do about it took a team of data scientists and engineers, the kind of function a large bank could staff and a company of two or three hundred people could not. Eva Narunovska's argument is that this is the thing AI has quietly changed. As chief executive and co-founder of Forse AI, based in Riga, she builds forecasting for mid sized firms that reaches a standard once reserved for the largest players, and she is blunt that the prediction itself is rarely the hard part.
She stresses the point that the model is rarely the hard part. Eva Narunovska has spent more than twenty years in data and analytics, including a stretch as a business intelligence programme manager at Swedbank. Most companies, she learned there, look only at what has already happened, examining various charts through a limited number of dimensions and then still relying on their own judgement about the future. Forse AI is her attempt to push past that, identifying patterns and relationships across hundreds of variables and their interactions simultaneously to move from what happened to predictive analytics - what is likely to happen, and then to the part clients actually want: decision intelligence - what to do about it. "I honestly love my job," she says, and it is easy to believe her.
What AI changes, in her account, is that a mid-sized firm can now reach the kind of prediction that used to be reserved for the largest players, and reach it without standing up a data-science function of its own. The raw material is usually already there. Those companies, she has found, tend to have more value locked in their data than they realise, scattered across CRM records, accounting files and internal systems, fragmented rather than absent. AI cannot simply compensate for poor data. The work often starts with bringing fragmented data together, understanding what it means in the context of that particular business and, where useful, enriching it with external signals., which is what lets a smaller firm utilise its own data the same way a much larger does.
What Forse AI hands back is not a single score but a chain. The platform will tell a client which leads are most likely to convert, which customers are drifting towards churn, where the is cross-sell or upsell potential and which action to take first when a hundred customers are all at risk at once. Crucially, it also explains why, for each individual case, a client is likely to leave or to buy. Narunovska is emphatic that this is not a nice-to-have. “Clients do not want predictions for the sake of predictions,” she says. What they want to know is what they and their people should do on Monday morning, and they will not act on a recommendation they cannot understand.
A prediction a manager cannot explain is often ignored, so Forse AI treats the reasoning behind each output as the value of the tool rather than a feature bolted onto it. It is also where domain knowledge stops being optional. The same field in the same CRM can mean opposite things at two companies. For example an agreement signature date - in one company signing means won; in another, a deal isn't truly considered won until payment because some signed agreements never convert into paying customers. A model that does not understand that context will confidently mislead.
She is clear that the techniques used by Forse AI are established data science rather than anything especially exotic, but each model is adjusted to the specific business rather than sold as a plug-and-play tool. Forse AI takes account of seasonality, identifies and accounts for anomalies that could otherwise distort the forecast,, and builds tailored solutions on the basis of the manager's own knowledge of how the business actually runs. The firm has also reached beyond the obvious, running a research collaboration with the University of Latvia on hybrid quantum and classical methods for forecasting, work that has since resulted in peer-reviewed research. It is a reminder that decision intelligence, the category she puts Forse AI in, is moving quickly.
The kind of prediction that once needed a bank's budget will keep spreading to mid-sized firms. Routine prioritisation will increasingly happen automatically, so that a sales rep knows who to contact, when to contact them and what to talk about. The decisions that carry weight, she is clear, stay with people. Forse AI's job, as she describes it, is to put the evidence in front of an experienced manager earlier than instinct would, and to explain it clearly enough to act on.
