The Chief AI Officer: Scientist, Architect, Coach — Rania Khalaf, WSO2
Rania Khalaf, Chief AI Officer at WSO2 and formerly a long-time leader at IBM Research, shares a framework for a role that is still being defined. She covers the Scientist, Architect and Coach focus areas, how the balance shifts by company type and AI maturity, and why she measures AI adoption, agent readiness and market signals instead of token counts. She also addresses the need for CEO backing and the question of who should own AI.
Three hats for a role that can become everything
AI now touches nearly every part of a business, which makes the Chief AI Officer role easy to stretch into everything. The approach that has worked is to split the job into three focus areas: Scientist, Architect and Coach. Each is better understood as a slider than a fixed title, and the right setting depends on three things: the type of company, its level of AI maturity, and the skills and background of the person in the seat.
The Scientist hat is about exploring, experimenting and building. Even at a low setting, the role does not require a PhD, but a team must keep experimenting because the technology moves too quickly to ignore. At the high end sits the creator-inventor. The Architect hat adds strategy. In a software company it means shaping product strategy and creating new software. In a company that does not sell software, it means using AI to cut costs and raise productivity while also opening new revenue streams and markets. This hat also carries board-level responsibility, since the role must report what AI is doing for the company and its customers. At the low end of the Architect slider is a steward who manages budgets, prioritizes projects by expected value, and reports results. That job depends more on business operating skill than on deep AI knowledge.
The Coach hat was the surprise. After twenty years at IBM Research, where she ran about a third of the global AI research organization, she found that evangelizing, educating and over-communicating take up much of the job. In companies that do not sell AI products, coaching runs inward. Teams need to discover limitations such as hallucination for themselves, because people do not believe it until they experience it. As a company matures, coaching shifts outward toward customers and the community. In her current role she acts as a trusted advisor to other CTOs and chief AI officers. She starts from the problems they face rather than from a pitch, and sometimes points them to other tools when WSO2’s products are not the right fit.
Mapping the framework to real careers
To test the framework, she had Claude and Gemini analyze resumes and job postings for the role and match them to the sliders. The results suggested that the role is defined mainly by the person, whose skill set determines where it lands.
Her own path illustrates this. At an agricultural biotech unicorn in the Cambridge area, which used CRISPR on corn, soy and wheat, validating a gene edit took ten years and a great deal of land. The first AI need was methods that help scientists form better hypotheses about which edits to make. The mandate grew quickly because the company had excellent geneticists but few engineers. She ended up building a data platform, moving to an all-AWS stack, adopting Databricks, and hiring data engineers and a CISO. One lesson stood out. A problem that seemed to call for AI, measuring corn embryo size as a predictor of editing success, was solved with simple blob detection, with no machine learning needed.
At WSO2 her split is roughly 20% scientist, 60% architect and 20% coach. The company has a small research group that produced three publications this year, two of them with undergraduates in Sri Lanka, one of which won a best paper award. With about 75% of employees technical and highly curious, inward coaching needs are low and outward coaching dominates. The strategy is now centered on the agentic enterprise.
What to measure, and why not tokens
When asked to measure AI usage, she refuses to track tokens because they are easy to game and say little about value. She tracks instead:
- AI fluency across the workforce: whether knowledge sits in a small center of excellence or whether builders exist across functions, with an AI lead in every product team.
- Depth of adoption: from simple tasks such as email cleanup or code completion to changing entire workflows, including agentic employees that work alongside teams.
- Tool availability: whether the company subscribes to tools and pays for team or enterprise accounts.
- GEO visibility: generative engine optimization, meaning whether LLMs can find the company and understand what it does. Content is increasingly written for agents, and the web is becoming hard for humans to read.
- Agent-consumable products: documentation that LLMs can consume, and every product exposing an MCP server, skills and a CLI.
- Agent-proof pricing: consumption-based pricing, so that agents cannot erode per-seat revenue, combined with dogfooding the company’s own stack and closing the feedback loop so products improve.
- Market and ecosystem signals: earned thought leadership such as invited talks, publications and press rather than paid articles, analyst recognition, customer adoption and use cases, community participation (stars and forks for a fully open source company), and AI ARR as new products mature.
Revenue is not the only number that matters, and the list should change as the company matures. As she puts it, you get what you measure, because everyone optimizes for whatever is tracked.
WSO2 and the agentic platform
WSO2 is about 20 years old, has roughly 150 million in ARR, and powers software in more than 90 countries across six continents. Its roots are in API management, integration, identity and access management, and internal developer platforms. Since she joined, the company has extended these into AI. It added an AI gateway to manage LLM interactions as APIs, agent identity that applies the rigor of human identity management to agents, and an agent builder in the integration platform. This year it released an agent platform that manages the full agent lifecycle, with an agent manager just launched. Everything the company builds is open source, not open core.
No silver bullet
The role needs strong CEO backing because it touches nearly every part of the company. Some AI-forward companies merge the chief product officer and chief AI officer into one person. In some non-software companies the head of HR has become head of AI, a development she admits she does not yet know how to feel about, even as agents are increasingly treated as part of the workforce. Her closing advice is to look for the intersection of what you are good at, what you love, what the world needs and what you can be paid for. Because the role is still fluid, shape it so that you sit at its center, or find a company whose alignment matches your interests, your strengths and your needs.