How to integrate AI into enterprise workflows using lean methods

Executive overview

Most enterprise AI initiatives fail not because of bad technology, but because teams skip user validation and jump straight to building. Nabeel Siddiqui, Global Senior Director at SAP, describes how his team applies lean startup methods to internal AI adoption — treating employees as customers and running structured experiments before writing a line of code.

The framework moves from hypothesis to experiment to pilot to production, with continuous user involvement at each stage to reduce resistance and improve fit.

Map the technology to the user's day-in-life before building anything.

Testing AI use cases before development

• Start with a hypothesis: e.g. "Can AI improve prospecting for our sales team?"
• Use Figma mockups as MVPs — run A/B tests with focus groups before any development
• Observe users rather than asking them; in AI, teams are often ahead of user intuition
• Use explainer videos to simulate the use case — "seeing is believing" prompts concrete feedback
• Run a concierge MVP: solve one problem for one user group, then expand
• Share pilot best practices broadly to crowdsource ideas from outside the focus group

Managing a portfolio of use cases

• Start with a wide list; validate which use cases get real user pull
• Persevere on use cases with clear demand; pivot when adoption doesn't materialise
• If a feature isn't used by one group, test whether it fits a different role or function
• Build reusable frameworks for sprints, focus groups, marketing, and business-to-technical translation — the "kitchen" any idea can be cooked in

Measuring AI impact

• Productivity is the primary KPI across most AI use cases
• Quantify time saved, then compound by total number of users
• For quota-carrying roles, saved time translates to pipeline and revenue
• KPIs form a hierarchy — productivity at the base, revenue and pipeline metrics above

Overcoming resistance to change

• Run 4–5 day sprints with focus groups; make end users co-designers, not recipients
• When users are decision-makers in the process, resistance drops and engagement rises
• Invite users to sketch solutions and bring outside inspiration — they become product contributors
• No ideas are bad ideas; celebrate volume of ideas and reward top contributors
• A culture of trials removes the ego risk from proposing new ideas

Driving adoption to the last mile

• Enablement is the critical last step — "build it and they will come" does not work
• Localise enablement sessions by region; include local representatives in the messaging
• Use a train-the-trainer model: one expert creates hundreds more through word of mouth
• Support with online learning modules and internal marketing to sustain adoption