Carrington SmurlProduct Designer / Engineer

Product Design2024

Adaptive Cards

Intapp was using the word "card" for two different things. Until that was fixed, no amount of redesign would have made the pattern consistent.

Role
Associate UX Designer, paired with a Senior Designer
Timeline
April – August 2024
Client
Intapp
Team
Diad with a Senior Designer; collaborated with the Design Systems team
Surfaces
DealCloud · Microsoft channels
Stack
Figma · FigJam · Pendo · Asana
Constraints
  • One pattern had to work across Intapp and Microsoft surfaces
  • Design- and technology-agnostic by requirement
  • Three separate user bases — Private Equity, Legal, Real Estate
  • Guideline changes needed Design Systems team approval
  • Handed off before visual design was finalised

The problem

Client feedback kept pointing at the same need: insights had to be shareable across Intapp and Microsoft product channels. Prior research had already concluded that a card redesign on the new v4 components was necessary, and specific use cases required cards to be broadcast to multiple channels as an Adaptive Card.

The design requirement was one sentence: a design- and technology-agnostic solution to unify the card pattern across multiple channels.

That is the kind of ask that can absorb months and produce nothing. So how do you tackle a large, ambiguous ask? That question — written on the board, in those words — was the real first deliverable.

And underneath it, found during the audit, was a problem nobody had named yet:

Right now we’re using the term “cards” for two different things: as containers and as actual UI elements. This is causing confusion and making it hard for us to talk about how we want our product to evolve.

Product managers, designers, and researchers were each using “card” to mean something different, then designing against their own definition. No component redesign would have fixed that. The next team would have reintroduced the ambiguity within a quarter.

What we inherited

A board titled How We Got Here — Background, Past Exploration, Card Experience Alignment — listing owner, designers and related initiative, with an outcome panel and two columns comparing a short-term and a long-term approach.
How we got hereA previous designer had done substantial work on mobile cards and started the framework we were asked to test for feasibility. Starting a project by writing down what already exists — who owned it, what it concluded, which approaches were already on the table — is how you avoid spending week one rediscovering week zero.

Framing before designing

A design approach diagram: a three-circle venn of Customers (user experience), Business (product management) and Technology (engineering), with desirability, viability and feasibility questions around it, above a two-column process showing Research — frame the problem, define, expectations — feeding Design — ideate, prototype, test.
The design approachDesirability, viability, feasibility, each with the question it actually has to answer: does this make sense for our users, can Intapp profit from it, can we technically build it. Underneath, the process split into research and design phases. For an ambiguous brief, agreeing the method is what makes the scope arguable instead of infinite.
A design fundamentals board covering responsive breakpoints for desktop, tablet and mobile, a five-Ws panel, and screen-resolution specifications.
Fundamentals, kept in viewAccessibility concerns. How will adaptive cards look across every platform? How does our layout change at these breakpoints? A pattern that has to survive both Intapp and Microsoft surfaces is a responsive problem before it is a visual one, so the breakpoints went on the wall early rather than being discovered late.
A colour-coded Gantt chart roadmap of the project's design phases across several months.
Design process roadmapHow often will we check in, how will we assign work, which tools will we use. The Senior Designer built this and we ran weekly check-ins against it, with Asana tracking how many tasks each piece actually took. Rough planning up front is cheap; discovering in July that nobody agreed on cadence is not.

Discover

Who it is for, and what they already do

Two boards side by side: Stakeholders and Roles listing design, engineering, product and executive groups, and a Stakeholder Map plotting people into keep satisfied, actively engage, monitor and keep informed quadrants.
Stakeholders and influenceRoles on the left, influence on the right — who to actively engage, who to keep satisfied, who merely to keep informed. On a change that would eventually require the Design Systems team to approve a rename, knowing in advance whose agreement was load-bearing is most of the politics solved.
A quantitative data board pulled from Pendo: browser and OS usage pie charts, feature-level usage trend graphs, screen-resolution distributions, and a list of logged product requests.
Quantitative baseline — PendoMy partner pulled the data mining in Pendo and the user feedback while I ran the competitive work. Screen-resolution distribution is the unglamorous chart that matters most here: it tells you which breakpoints are real and which are hypothetical, and that decides how many card variants have to exist.

Competitive and comparative analysis

A large competitive and comparative analysis board: a comparison matrix across many products above a wall of annotated card screenshots from competing applications.
Competitive / comparative analysisDirect and indirect competitors, pulled in with their card patterns and annotated against heuristics. The question I kept returning to is the boring one that actually mattered — what separates a card from a container, from a tile on a dashboard? Without an answer, “unify the card pattern” has no test for success.
A board of direct-competitor screenshots annotated with magenta callouts marking card regions and interactions.
Direct competitors, annotatedAnnotating rather than collecting. A screenshot wall proves you looked; the callouts are what turn it into an argument about which conventions are shared and which are one company's habit.
A board titled Microsoft Dynamics — screen sizes and cards, showing the same interface captured at three widths.
Microsoft Dynamics at three widthsMicrosoft got its own study, because this pattern had to live inside Microsoft channels as well as ours. Designing a component that will be a guest in someone else's product means their conventions are a constraint, not a reference.

Our own current state

An annotated analysis of the existing product's card interactions, with magenta callouts attached to each interactive region across several screens.
Current card states and interactionsAn inventory of every existing card state and interaction before proposing anything. Unglamorous, and the only way a recommendation survives the question “did you look at what we already have?”
A specification board documenting current DealCloud card functionality — overview, card states including default, selected and focus, and hover state — illustrated with example cards.
Spec of current functionalityDefault, selected, focus, hover — each state written down with the behaviour attached. Before you can agree what a card should be, somebody has to be able to say what it currently is, and that document did not exist.

Define: the word was the problem

A FigJam board titled Taxonomy Definitions, comparing how Intapp, Microsoft, IBM, Google, Hubspot and Salesforce each define a card, with annotated sticky notes and a diagram labelling window, frame, panel, tile, card and toolbar.
Taxonomy definitionsDesign systems laid side by side — Intapp, Microsoft, IBM, Google, Hubspot, Salesforce — each with its own definition of a card, highlighted for where they agree. The diagram separates window, frame, panel, tile, card and toolbar, because half the disagreement was people naming different objects with the same word. This board is the argument that the rename was necessary, which is what made it survivable with stakeholders.

The recommendation was to split the overloaded term in two:

  • Container — the old “card.” A wrapper that consolidates related information into a unified interface. Organises and groups; minimal interaction beyond expanding or collapsing.
  • UI card — net new. A standalone component presenting concise, actionable information, or acting as an entry point into deeper functionality. Interactive by definition.

I wanted to be certain the recommendations would still be appropriate once cards were renamed to containers and “card” became a net-new thing — which is a harder test than it sounds, because it means every existing usage in the system has to resolve cleanly under the new names.

Define: three user bases, one component

To understand how cards should behave on a dashboard, I worked through the archetypes across Intapp’s three user bases — Private Equity, Legal, and Real Estate. I pulled from prior research documentation and the ACTIVATOR persona our research team had validated earlier that year, went to research office hours with questions, and spoke with internal subject-matter experts to pressure-test my assumptions.

I also reached out through my own network for an interview with a private equity user, because the internal picture had gaps I could not close from documentation.

A board titled User Stories — By Outcome, with three colour-coded columns of user stories for Real Estate, Private Equity and Legal, each story paired by an arrow with a concrete UI idea, beside a column of interview notes from a private equity subject-matter expert.
User stories → UI ideasEvery story is paired with the card idea it implies — “as a legal assistant scheduling appointments, I want a user-friendly interface…” becomes “card-based layout for calendar entries.” The orange column on the left is the raw interview with the private-equity SME I found through my own network. Mapping story to idea in one artifact is what stops research becoming a document nobody opens.
Three empathy maps side by side, each built around a persona avatar with branching sticky notes.
Three empathy mapsEmpathy mapping narrowed a sprawling set of possibilities down to three use cases worth solving. Narrowing is the deliverable here — an agnostic pattern that tries to serve every case serves none of them.
An End User board laying out the three selected use cases as grouped sets of sticky-note cards.
The three use casesThe three we committed to solving, written out as end-user situations rather than features. Everything downstream — the pattern flow, the variants, the principles — gets tested against these three and nothing else.
Two large flow diagrams mapping sequences of screens and decision points across the card experience.
Flows behind the use casesEach use case taken through to the sequence of screens and decisions it implies. This is where an abstract requirement becomes testable: a pattern either carries these flows or it does not.

The rule that outlives the project

A large decision-tree flowchart titled Pattern Flow, with diamond decision nodes routing to platform card or adaptive card outcomes.
Pattern flow — when to use whichA decision tree for when to use a platform card and when to use an adaptive card. This is the artifact I would point at if asked what I actually shipped. A taxonomy tells people what the words mean; a decision tree tells them what to do on a Tuesday, which is the thing that stops the ambiguity coming back once everyone who made the decision has moved on.
A written project-overview document with sections for description, use cases including main and alternate flows, and user stories with outcomes.
Documentation, off the FigJam boardGetting everything off FigJam. Before wireframing we laid out the ground rules for the Figma file for everyone who does not live in Figma like we do — description, use cases with main and alternate flows, user stories with outcomes. Research that only exists as sticky notes has a shelf life of about one reorg.
A design principles document with tracked edits visible, setting out three principles — identifiable, informational, actionable — each with supporting bullets about card composition.
Design principles, mid-editCard content should be identifiable, informational, and actionable. The strikethroughs are deliberately left in: these guidelines covered both the new and the renamed components and required Design Systems team approval, so I drafted them in collaboration with that team rather than handing over a finished document to be argued with.

Wireframing

Three cards side by side: an older v3-style card with description, industry sector and employee-count fields and fit chips, next to two new cards with a title, image, repeated status rows and a primary email action.
Before and after — the component itselfThe existing card on the left, the proposed replacements on the right. The change is not decoration: the old card is a read-only summary, and the new one ends in an action. That is the taxonomy made visible — a container holds information, a UI card does something.
A wide WIP analysis strip showing six relationship-card variants — horizontal and vertical, with and without relationship data and relationship intelligence, plus collapsed — each annotated with header, body and footer labels.
Variant matrix — WIP analysisEvery combination that has to exist: horizontal and vertical, with and without relationship data, with and without relationship intelligence, and collapsed. Each one labelled header, body, footer, so the anatomy stays constant while the content varies. This is what “technology-agnostic” actually costs — the pattern has to hold its shape through all of it.
Final wireframes for Real Estate cards in four density states — expanded with labels on some fields, on all fields, with no labels, and collapsed — alongside relationship cards with an actions menu and v3 versus v4 colour swatches.
Final wireframesReal estate and relationship cards at four densities, from fully labelled to collapsed to a single line of text. Labels are the variable because they are what a scanning user does not need and a new user cannot do without — and the v3 and v4 colour swatches sit beside them as the reminder that this all has to land inside an existing system.

Handing off

In August the work was handed to the Design Systems team to finalise the visual layer.

The outcome was still in progress when I handed off, and I am not going to claim a result I did not see. What shipped out of my half: a taxonomy the organisation agreed on, the competitive analysis behind it, a quantitative baseline, three validated use cases with flows, a pattern flow, rewritten guidelines approved by the team that owns the system, and a variant matrix specified down to anatomy.

What I would tell you in an interview

The lesson I took is that design system work is frequently language work. The ask was a component redesign. The actual blocker was a word carrying two meanings, and no component would have resolved it.

The second lesson is about what survives you. Of everything on this page, the artifact with the longest life is the pattern flow — the decision tree that answers “which one do I use here?” without anyone having to find me. Research boards get archived. Wireframes get redrawn. A rule people can follow is the part that stays.

What I would do differently: I would have pushed for a usage metric before handoff. I have the qualitative case for the split and the stakeholder agreement, but no measurement of whether the new taxonomy reduced the confusion it was meant to reduce — and without that, the next person to question the rename has nothing to check it against.