Building a horizontal framework with AI
- Role
-
Lead Experience Designer
@Canonical ⊂ 2024 -
Canonical has a wide portfolio of over 60 deeply technical products. After I joined, I recognized that this breadth makes it increasingly challenging to have meaningful conversations about target user segments across marketing, sales, product, design, and engineering. There was a good understanding of the audience in the product and engineering teams through community engagement and plenty of dogfooding, but transmitting learning between teams was where things broke down. This led to friction both within a single product’s experience and cross-product experiences in the portfolio, for example sales not targeting the right type of customers or mismatch in touchpoints across user journeys.
A job map for a core user segment with the JTBD framework
From one segment to a framework
I decided to change this, and to take on the challenge to improve the business. As a first exploratory step, I worked with the marketing team to describe a single segment of our audience using the JTBD framework. I’ve led several workshops with senior stakeholders to gather data. Besides the concrete benefit, the goal was to understand the problem and pitch a solution to leadership. The first round was successful enough that we expanded it into a proper project team to create a framework for both positioning products and measuring their performance using a mix of customer journey, jobs to be done, and persona methods. The framework was intended to provide common terminology and a set of artifacts that could drive conversations across silos, not just within design.
A framework to align teams
This project could have grown into something very expansive, so we searched for a key leverage point. During our exploratory conversations with stakeholders across 10+ teams, we found out that a shared description of the target audience would help the widest range of teams. Sales could better understand the customer teams they were working with, marketing could plan campaigns more precisely, product managers could write more targeted content, engineering could improve their documentation, and design would have clearer design targets.
Archetype example samples
A two-tier system
With this in mind, we created a set of archetypes. I designed a two-tier system: a strategic tier with cross-portfolio archetypes relevant for sales and marketing efforts spanning multiple products, and a tactical tier with product-level personas offering vertical depth for product teams.
This had to be a system that evolves as teams learn, not a set of static artifacts that go stale.
Keeping the system alive
Personas tend to decay as teams accumulate new information that never makes it back into the original documents. To solve this, I designed an agentic workflow that continuously synthesizes new information from across the organization, drafts updated archetypes, and validates them against their sources, so the written knowledge stays current instead of going stale.
That durability turned out to matter beyond the archetypes themselves. Because the segments now live as structured, well-maintained written knowledge, they’ve become reliable context for other agentic workflows across the company, a shared foundation teams can point their own agents at, rather than each rebuilding an understanding of our users from scratch.
The agentic workflow that keeps the archetypes current
Adoption across the org
- 60+
- products in one portfolio
- 10+ teams
- aligned through discovery
- Sales & marketing
- adopted the archetypes
The new archetypes were immediately adopted by the sales team, we helped them by designing new onboarding and training artifacts. The marketing team used them to update their battlecards and bring more direction to messaging workshops and content plans. Over time, we also ran workshops with product teams using the tools we designed for them to work with the archetypes in their own context. The most interesting outcome wasn’t any single artifact. It was that teams across the organization now had a shared language for talking about users, and a system that keeps that language grounded as their understanding grows.