Two fresh resources from Stanford’s SALT Lab show a simple truth: most wins come when you build what workers actually want. Here’s how to use that insight to pick better use cases and get to production with IAaaS.

What the new research says

The SALT Lab team built a worker-centric audit of AI at work. They asked 1,500 people across 104 occupations which tasks they want an AI agent to take on, and compared that with what experts think is technically feasible. They also introduced a Human Agency Scale (H1–H5) that describes how much human involvement a task should keep. H3—equal partnership between human and AI—shows up as the dominant preference in nearly half of occupations. In other words, most people want help, not handoff. [Paper PDF] [Project site]

The study maps tasks into four zones: “Green Light” where workers want automation and the tech is ready, “Red Light” where the tech exists but workers do not want automation, an R&D zone where workers want help but today’s tools fall short, and a Low Priority zone. One finding stands out: a lot of current investment and attention sits in the Red Light and Low Priority zones, while Green Light opportunities are underused. That misalignment explains why so many pilots stall.

What this means in practice

You will move faster if you start in the Green Light zone. That is usually high-volume back-office work with clear owners and measurable outcomes. Think invoice intake and coding, vendor updates, order changes and RMAs, contract or policy lookups across large document sets, and shared inbox triage.

Use the Human Agency Scale to set how your solution should behave. Many good fits are H2–H3. That means the AI should do the reading and the heavy lifting, but hand off to people at clear points. Confidence thresholds, exception queues, and one simple rule—when in doubt, send to a human—go a long way.

A simple delivery playbook

Line up the business first. Run a short intake with each team, keep ideas in one tracker, and rank them with a single scorecard: value, time to live, data readiness, and a named owner. Agree on two success measures before you start.

Build with a small starter kit so every solution behaves the same. Use document intelligence and retrieval-augmented search to read emails and PDFs. Apply rules and thresholds for decisions. Let digital workers handle the clicks and updates across your apps and the web. Keep data in place. Deploy inside your tenant, use your identity controls, and log every action.

Why IAaaS helps

You can do this in-house. Many teams can. Most do not have the time. With Intelligent Automation as a Service, Optezo designs and builds the automations, deploys inside your tenant, and runs them day to day with SLAs, monitoring, and updates. You focus on the business. We keep the digital workers productive.

If you want broader context, see our posts on Hyperautomation 2025, Responsible Automation, and From 2 Bots to 20.

Bottom line

Build where workers say “yes” and the tech is ready. Start in operations, set the right level of human involvement, keep data in your tenant, and measure what matters. That is how you move from pilots to production.

Ready to ship Green Light work this quarter?