The dangerous bit looked like a dropdown
A Quickbase builder can authorise Slack, Shopify, QuickBooks, Gmail, Snowflake, or an internal data source, then let Pipelines run against it without being a developer. The risky parts are rarely dressed like risky parts. Sometimes they are a sandbox-or-production selector tucked into a connection form.
I worked on Pipelines for three years: reusable OAuth anatomy, AI-generated steps, Bulk Trigger, troubleshooting, and Jinja code mode. At first I treated trust mostly as an explanation problem: make the flow clear enough and people will understand it. The work kept correcting that. Trust also came from where the product stopped, what it refused to commit, and whether a builder could inspect a consequence before it touched real data.
My part in it
- Role: Lead Product Designer across three years of Pipelines work.
- Scope: reusable connection patterns, AI-assisted creation, high-volume automation, troubleshooting, and expert configuration.
- Product rule: AI could propose and populate work, but people could inspect and change it before a run.
- Strongest measured reach: from September 11, 2025 through September 11, 2026, 2,826 customer accounts used the all-Pipelines-AI feature set; 1,061 used Generate Steps with AI.
- Evidence boundary: the figures below are product-analytics counts and action sequences. They show reach or follow-through, not completed automations, revenue, or causal credit for my design.
Connecting an account without hiding the dangerous choice
Every provider asks for its own peculiar thing: OAuth scopes, store URLs, sandbox or production choice, service accounts, or a permissions ladder long enough to make anyone click through on instinct.
Designing a bespoke screen for each connector would have made every new integration a fresh argument. I built one connection anatomy and gave provider-specific complexity a defined place inside it. Permissions were written in plain language instead of raw scope strings, because the person authorising a business workflow is often not an OAuth engineer. Sandbox versus production became a first-class choice, because discovering an environment model after writing to the real books is a terrible onboarding sequence.
The point was not to make every connector look identical. It was to make the decisions that could damage data behave consistently: what account is being connected, what access is being granted, what environment is selected, and what the builder can do next.
Letting AI suggest work, then making it wait
The AI creation work began with a practical problem. A person could know the business action they wanted but not the sequence of pipeline steps, field mappings, or syntax needed to build it. The model could help with that scaffolding. It should not get to silently decide that the scaffolding was correct.
The interaction let a builder describe a step, review a proposed sequence, and inspect populated fields. A proposal remained editable. The underlying state model had to account for loading, no answer, usage limits, timeouts, permissions, and recovery, which are usually absent from a pleasant demo until they are suddenly all the product has.
Pendo counts from September 11, 2025 through September 11, 2026 show 2,826 customer accounts using the all-Pipelines-AI umbrella feature and 1,061 using Generate Steps with AI. Those figures are not additive: Generate Steps is one part of the broader AI surface.
Monthly umbrella reach moved from 57 accounts in August 2024 to 862 in August 2026. That is observed reach across an evolving feature set, not a claim that one design caused the growth. The headline creation action moved in the other direction: Generate Steps reached 173 accounts in April 2025 and 114 in August 2026. The product had more AI surfaces by then; the data does not show whether builders migrated, succeeded, or preferred one path.
A useful funnel still is not an automation outcome
Smart Builder gave people a way to begin from a request and open the generated pipeline. In the measured visitor sequence from June 13 through September 11, 2026, 42% went from opening Smart Builder through building and populating steps to opening the generated pipeline.
That final action does not prove that the automation worked. It only shows that someone followed the measured sequence to inspect the pipeline. I left the distinction in because it is exactly the kind of conversion number that can become a portfolio victory lap if nobody asks what it actually records.
There is another limit on the story. On August 17, 2026, the dated project record says Smart Builder was hidden when the Pipelines agent orchestrator was on. That establishes that the older flow was being superseded in that configuration. It does not establish a future roadmap, a migration story, or whether the replacement is better for a particular builder.
Designing for thousands of records
Pipelines originally fired one row at a time. That model gets silly when somebody needs to move thousands of records.
I worked on the Bulk Trigger UX and its feedback loop, including naming cleanup. “Bulk” and “Batch” had been used loosely enough that people could discuss the same feature and mean different behaviour. The terminology work was boring, slightly annoying, and load-bearing.
In the July 20, 2025 through July 19, 2026 Pendo window, the archive records 60 accounts using the On New Bulk Event Trigger and no detectable adoption lift after the redesign. That is not the success story I would have preferred. It is the useful one: a niche, high-consequence capability needs a denominator and a definition of a good run before its product effect can be judged. Engineering owned the performance improvements, so I do not attribute those results to my design work.
Making room for advanced builders
The visual editor made Pipelines approachable. It also became a ceiling for people who knew exactly what they wanted to write.
For power users, I designed a Jinja code-mode editor and the lock and disable rules needed so code mode and visual mode would not corrupt each other. When code mode owns an expression, visual controls have to make that ownership visible rather than quietly rewrite it.
Supporting both paths created ugly rules. It doubled a class of edge cases. Removing code mode would have made the product tidier by frustrating the builders most likely to hit its limits. I chose the uglier state model because the alternative hid a real capability boundary.
Troubleshooting made the evidence less flattering
Troubleshooting reached 969 accounts from September 11, 2025 through September 11, 2026. Of 164 feedback events, 35.4% were positive. That is a limited feedback sample compared with the feature’s overall use, so I treat it as a directional sentiment signal, not customer satisfaction.
The negative signal belongs in the case study because it changes how I read the rest of the work. A reusable OAuth pattern, editable AI proposals, and specialist escape hatches can make a complex system legible. They do not guarantee that a builder understands why an automation failed or can repair it. I can show product reach and a measured path to opening a generated pipeline. I cannot show a clean success rate, unprompted confidence, or a causal result from the redesigns.
What three years on one domain taught me
The OAuth-anatomy bet is the one I would make again. A parameterised pattern felt slow when the first connector could have shipped as a one-off screen, but it gave later integrations a stable place for the decisions that mattered.
The broader lesson is less elegant. AI may propose work; the interface still has to show what will run, what has changed, which representation owns the value, and where a human can intervene. I learned most of that after the happy path already looked finished. The remaining gap is measurement: the product needs a way to distinguish a pipeline being opened from an automation being useful.
Related work
Designing Pave shows how the same inspectability question shaped a public AI app builder.
Untangling Connection Central follows connected resources and pipelines through a research-led information architecture.