Taking AI From Pilot to Production Scale

AI usage is surging among workers, and it’s having an effect on individual productivity, with 64 percent of respondents to Salesforce’s Slack Workforce Index reporting they were more productive.

Yet more than 80 percent of executives surveyed by McKinsey in July 2024 said generative AI had yet to move enterprise-level earnings in a tangible way. And Gartner this year predicted that over 40 percent of agentic AI projects will be canceled by the end of 2027, owing to escalating costs, unclear business value or inadequate risk controls. According to an April 2025 survey from Capgemini, just 2 percent of organizations have deployed AI agents at “full” scale, while most projects remain at the exploratory stage. Fewer than one in five organizations surveyed reported high maturity in the data and infrastructure needed.

The reason pilots fail to hit production scale is simple: AI is not plug and play. Before you can get agentic AI to scale, you need maturity in people (capabilities), processes (ways of working), and data (a single source of truth). As Ethan Mollick, a professor at the Wharton School of the University of Pennsylvania, puts it: “AI use that boosts individual performance does not naturally translate to improving organizational performance. To get organizational gains requires organizational innovation.”

Building a successful AI project

The cost of experimentation is near zero with AI – any team can spin up a pilot – but the gap to production scale and actually having an impact on your bottom line has increased.

Because EQT invests across the stack – from early-stage venture to mature buyouts – we can see what works across the portfolio and apply lessons to other businesses facing similar problems.

In our experience, successful AI projects:

1. Have clarity of purpose – They connect any AI investment to strategy and business outcomes, and business and technical teams understand the purpose of the project.
2. Target enduring challenges – They focus on problems that are worth solving.
3. Focus on the problem, not the technology – They use AI where it is the right solution.
4. Invest early in data – They invest in the underlying data infrastructure and data governance, so that they can take use cases to production with speed and quality. To fulfill their full potential, agents need access to fast streams of high-value, timely data via well-documented APIs.

EQT frames the AI opportunity in three ways to help our portfolio companies structure their strategy:

1. Productivity – Internal efficiency to cut costs and/or deliver faster.
2. Innovation – Product/service development to deliver customer value.
3. Competitive shifts – Building a defensive moat, and keeping pace with the competitive landscape and potential disruption.

A digital transformation is the process of developing organizational and technology-based capabilities that allow a company to continuously improve its value proposition, lower its unit costs, and, over time, sustain a competitive advantage. As Bill Conner, CEO of AI architecture provider Jitterbit, puts it, “Success with AI is less about bold promises and more about disciplined integration.”

Because of this, we often address our talks on AI to one underserved audience: CFOs. They are the ones who can challenge a technical team to make sure a pilot has a path to commercial impact. CFOs can tie a project to business strategy, ensuring a viable route to scale and that outcomes will show up in the bottom line. It’s also why EQT is adding digital transformation champions to our boards and forming digital committees to pressure‑test roadmaps, set ethical guardrails and monitor return on investment.

Meeting companies where they are

This is a long-term practice, not a one-shot thing. We need to meet companies where they are and then help deploy capital in the right areas to bring them up to a level where they can reap the benefits of AI and future-proof themselves to adopt the next generation of technology. Depending on the maturity of the company’s systems and the use case, we apply the right technology (ML, GenAI, agentic solutions) after getting the basics right first.

Here are a couple of examples from our portfolio.

With SK Shieldus, a Korean security company, we first provided a consolidated view of customer information via Databricks and improved the churn scoring model with AI. With that, we found the customers that were at risk of leaving and extended that cohort’s lifetime significantly. That corresponded to a meaningful percentage of revenue retention impact that fell straight down to the bottom line, several million dollars on a yearly basis.

With Credila, an Indian provider of education loans, we started by modernizing the core loan origination system and loan management system. On top of that solid foundation, we are now building AI solutions to improve fraud detection and the efficiency of loan applications, reducing time by an order of magnitude. This would not have been possible if they hadn’t had the basics in place first.

In both cases, the pattern was the same:

1. There was a clear business problem to solve.
2. They invested in the right data infrastructure first.
3. AI was the right solution for the problem, sometimes starting with a simpler ML solution.
4. They shipped a minimum viable product (MVP) first and solved one job end-to-end, before layering on extra features.

It is easy to overestimate the short-term benefits of AI in a lower maturity company and to underestimate the long-term impact. AI is a digital maturity game. That is how you get AI in production at scale. It’s a journey of building maturity.

Download EQT's full report on The AI Opportunity here.