Repsol hired Designit to create a company-wide playbook for how employees should experience generative AI. The result, described by Designit as a human‑centered foundation spanning tools from Microsoft Teams to embedded platforms, aims to replace scattered pilots with a single, consistent approach to AI at work (Designit).
That move is less about models and more about trust. In large organizations, the hardest part of AI isn’t the algorithm. It’s giving thousands of people a coherent, predictable way to use it. The Repsol Designit AI framework addresses that gap head‑on.
Inside the Repsol Designit AI framework: one playbook, many contexts
Designit says research inside Repsol found strong demand and plenty of experimentation, but also fragmented experiences. Interfaces didn’t match. Expectations varied. Confidence wobbled as employees tried to learn new patterns on the fly. From those findings, the partners defined principles to guide every interaction with AI: set clear expectations about the system’s role and limits; be transparent; invite feedback; make errors easy to spot and recover from; and bake privacy into each step (Designit).
The framework flexes by audience and setting. A bot inside Teams needs different cues than a model embedded in a line‑of‑business dashboard, but both should feel familiar and safe to use. A shared tone of voice and interaction style ties them together so the experience travels with the user, not just the app.
This emphasis lines up with outside guidance on trust and transparency. Usability researchers have urged teams to explain what an AI can and cannot do, reveal uncertainty when it matters, and support recovery from bad outputs, because that is how confidence forms in real use (Nielsen Norman Group). The Repsol Designit AI framework stitches those ideas into one operational system.
Why a unified AI experience beats one‑off tools
The lesson here is simple: people adopt patterns, not features. A dozen isolated assistants each teaching their own rules slow learning and seed mistrust. A single experience language speeds it up. That’s the same reason design systems reshaped web and mobile software a decade ago. Once components and behaviors stabilized, teams shipped faster and users stopped relearning the basics every time they opened a new app.
Generative AI raises the stakes because it introduces uncertainty by design. Outputs are variable. Explanations can be fuzzy. Without patterns that set expectations and create guardrails, teams risk chaotic rollouts and shadow usage. The Repsol Designit AI framework tries to turn that uncertainty into a managed, teachable rhythm across the company.
There’s a governance angle too. As rules tighten, organizations will need not just model evaluations, but interface standards that show how risk is handled at the point of use. The European Commission has highlighted transparency and human oversight as core to responsible deployment, which inevitably shows up in UI choices and interaction flows (European Commission). A repeatable AI experience framework makes that visible and auditable.
What this AI experience framework changes for developers and product owners
A consistent playbook reshapes day‑to‑day work for builders. Product teams can stop debating first‑principle questions on every project and start from shared defaults: how to disclose model limits, when to show confidence cues, where to place privacy prompts, and which fallback paths to offer when generations miss the mark. That cuts design churn and speeds code reuse.
For developers, it also clarifies handoffs. If transparency requires a short “how it works” note or a quick way to compare the model’s draft to the source, those patterns become components, not one‑off hacks. If error recovery means saving prompts and exposing edit history, teams implement it the same way each time. Over time, these patterns become testable quality gates rather than subjective debates.
Security and compliance teams gain something practical as well: a surface they can review. Privacy‑by‑design stops being a slogan when specific UI steps—consent notices, data retention status, export options—are standardized and enforced. Guidance from responsible AI toolkits can flow into those patterns so teams don’t reinvent policy in code on every sprint (Microsoft).
- Fewer new patterns to learn, which speeds employee confidence and reduces support tickets.
- Shared components for transparency and recovery, which shorten delivery cycles.
- Visible privacy steps, which make audits and internal reviews simpler.
What to watch as Repsol scales the approach
The concept is sound. The hard part is keeping it alive as tools multiply. Two tests will matter. First, whether the principles hold under pressure when pilots move into critical workflows and model behavior meets edge cases. Second, whether teams add new interaction patterns only when the principles demand it, not because a vendor ships a shiny feature.
If Repsol’s internal feedback loops play out the way the partners intend, the Repsol Designit AI framework should grow by evidence, not anecdote. That means tracking which transparency cues reduce bad handoffs, which recovery paths actually get used, and where privacy prompts cause friction. The best design systems evolve with this kind of telemetry. AI needs the same discipline.
There’s also the cultural piece. Designit describes interviewing employees and mapping tools to understand how people were already bending AI to their work. Keeping that discovery muscle active will matter as new roles—prompt authors, reviewers, domain curators—emerge inside teams (Designit). When those roles shift, the experience rules should shift with them.
If the approach holds, the signal for other enterprises is clear: a company can deploy many models, but it should teach only one way to use them. That’s what the Repsol Designit AI framework tries to codify—trust, clarity, and control expressed in interface decisions, repeated across every tool an employee touches. For more on this, see bloomberg.com.