BackEngine

The design challenge

An AI insight is only useful
if someone trusts it enough to act.

01 / Make the next action clear

Enough context to judge.
Enough structure to follow through.

The signal card was a design problem of its own. We iterated on how much to show before a user opens the detail: a concise summary, sentiment, importance, a source quote, and attribution. The detail opens over the feed, keeping the triage context in place while highlighted passages expose the evidence. Assignment, response, and resolution belong to the signal itself. Saved filters and “Assigned to me” turn the same feed into a personal work queue; creating a Ticket or Situation carries the evidence into the next step.

02 / Start with the job, then offer the controls

Different roles.
A shared understanding of the customer.

An account manager needs to know what needs attention today. A product leader needs to see which problems keep recurring. We designed role‑specific views over the same signals and organized chart presets around Product, Support, Commercial, and Market questions. Users can start with a relevant question instead of constructing an analysis from a blank screen. The shared data model keeps operational detail and strategic analysis connected.

03 / Let a useful starting point become a precise tool

Make the comparison
visible as it takes shape.

A preset is a starting point, not the limit of the workflow. Each chart line has its own filters, so a user can compare customer groups, sentiment, or signal categories in one view. We kept a live preview beside the controls: users can see what their choices produce before adding the chart. Wide and square formats make dashboard composition a simple choice, while each person’s layout stays independent of their teammates’ views.

04 / Carry trust into every level of analysis

A bigger conclusion still needs
a route back to the source.

We extended traceability from individual cards to AI‑generated reports. Each report conclusion can open the signals behind it, while a source‑volume summary makes the scope of the analysis visible. Free‑form questions support exploration; curated reports package expert prompts for more specific decisions and expose those prompts before generation. Both use the same evidence and version‑history model. Users can investigate a new question without learning a different way to verify the answer.

05 / Preserve context when the question gets specific

Prepare for the account.
Keep the evidence close.

A renewal conversation demands a customer‑level view. We brought sentiment, revenue, renewal timing, contacts, and communication history into one account workspace. A user can move from the overview into messages and the signals derived from them, or open a filtered view in the global feed. That connection works in both directions: start with a risk and inspect its source, or start with a conversation and see what it revealed.

06 / Design for where the work happens

The insight should reach
the people who need it.

Useful intelligence cannot depend on everyone remembering to open a dashboard. We designed recurring email and Slack subscriptions for signals, charts, themes, and reports. Content is re‑evaluated against current data at each run, with recipients and cadence configured for each channel. Together with ownership and workitems, this closes the interaction loop: discover an issue, understand it, give it an owner, and keep the relevant people informed.