Orchard - AI Skills Manager

The role
This is a new kind of role: part knowledge engineer, part workflow builder, part internal product owner. You will sit behind our Transformation Lead operationally, taking requirements from across the business and turning them into production-ready AI skills that the wider team can reach for on every project.
You are not just a prompt engineer. You will understand how the whole system works: how skills connect, how Claude's context window and tool use interact, how to structure knowledge so it can be retrieved reliably. And you will constantly be asking whether what we have is the most efficient way to do it.
You report to the Transformation Lead and work closely with discipline leads across strategy, creative, accounts, and development. You will also interface with the broader Enero Group AI community as Orchard's skill-building capability becomes a reference point for the group.
What you'll do
Skill writing and production
This is the core of the role.
Write, test, and iterate AI skills: system prompts, workflow logic, tool configurations, and usage documentation, to a high production standard
Take requirements from the Transformation Lead and discipline leads, translate them into functional skill specs, build the skill, and validate it works as intended before release
Join existing skills together into coherent, efficient systems: not just individual prompts but end-to-end workflows
Constantly review what exists and ask: is this the right way to do it? Is this efficient? Does this still reflect how we actually work?
Iterate based on real-world use, fixing failure modes, tightening outputs, and updating skills when models or agency processes change
Knowledge management and infrastructure
Own the AI skills library: structure, versioning, naming conventions, metadata, and access, so every skill is findable and usable across the agency
Manage the skills repository (GitHub or equivalent), including version control, branching, and review processes
Help build Orchard's organisational memory by documenting how we work, how our knowledge is structured, and where things live, in a form the AI system can actually use
Run regular audits to retire outdated skills, flag gaps, and maintain quality standards across the library
Quality, testing and performance
Build and maintain testing frameworks to validate skills before release
Establish and uphold the release standard: the bar every skill must clear before it goes into use
Track usage, adoption, and impact across the agency; surface patterns, gaps, and failure modes to the Transformation Lead
Maintain a structured feedback loop with teams: discovery before building, ongoing input after deployment
Best practice and standards
Keep pace with advances in prompt engineering, agentic design, and model behaviour, and bring new thinking back to the agency
Maintain Orchard's internal AI skills best practice guide
Embed responsible AI principles into every skill: bias checks, output validation, and appropriate human oversight
What you bring
Must have
Genuine fluency with large language models: you have built, tested, and refined prompts professionally, understand how context windows work, and can reason about why a skill is or isn't performing
Strong systems thinking: you see how workflows connect, where failure points live, and how to design for reuse and scale rather than one-off use
Experience managing a structured asset repository (GitHub or similar), where version control, branching, and review processes are second nature
Clear, precise writing: you can document a skill so that someone who wasn't in the room can pick it up and use it correctly
Comfort working across technical and non-technical stakeholders, translating between what teams need and what the system can do
A background at the intersection of technology and business: information systems, digital operations, knowledge management, or a similar discipline where you've had to understand IT systems from a business perspective
Highly valued
Familiarity with agentic AI frameworks: MCP, tool use, multi-step pipelines, or Claude's agent SDK
Experience building or maintaining internal knowledge systems, prompt libraries, or documentation platforms
Basic scripting or automation skills (Python, JavaScript) for testing pipelines and tooling
Background in a creative, digital, marketing, or production environment where AI tooling was part of the delivery stack
Exposure to healthcare or pharmaceutical marketing
You are
Obsessively quality-conscious. You'd rather ship one skill that works perfectly than five that kind of work
Genuinely curious about AI. You follow developments, test new models, and bring new thinking back to the team
Comfortable with ambiguity. This function is still being built, and you have real influence over what "good" looks like
Patient and methodical. You understand that getting teams to actually use something takes as much care as building it

Employment type
Full-time

Work arrangement
Hybrid