From 1,200 Calls to a $1.3B SaaS

The CIO had a whiteboard full of boxes and arrows. None of it matched what people actually did.

On the call, they kept coming back to the same problem: leadership was being asked to find “automation savings,” but nobody could clearly explain the day-to-day steps behind the work. The real process lived in a few experts’ heads and a scatter of docs in old folders.

Jennifer Smith wasn’t pitching anything. She was pressing for specifics: “How do you train someone new?” “Where is this written down?” “Who knows the real steps?” The answers were messy, and consistent.

That conversation was not special. It was one of roughly 1,200 interviews she ran over about three years with CIOs, CTOs, VPs of IT, and chief digital officers while she was at Greylock Partners.

Later, she would build Scribe, a tool that helps teams turn real work into step-by-step guides, and it has been used to document over 10 million workflows. This article is about the part before the product: how to run discovery with enough volume and structure that you stop guessing.

If you are a founder who is still choosing what to build, the practical takeaway is simple: run interviews like an operational machine for 4–6 weeks, look for repeated workflow pain (how work gets done), then ship a rough test and measure pull by behavior, not compliments.

Turning a VC Seat into a 1,200-Conversation Lab

Most founders hear “do customer interviews” and treat it like a side task. Smith treated it like production.

Over about three years at Greylock, she used a repeatable set of questions across roughly 1,200 conversations. The goal was not to collect opinions. It was to surface patterns in how big companies actually operate, especially where workflows break down.

Her interview targets were consistent: CIOs, CTOs, VPs of IT, chief digital officers, and similar leaders responsible for systems and operations. She asked what was breaking, what was slowing teams down, and what they wished Silicon Valley would build.

From Manual Slide Decks to Auto-Captured Workflows

Smith had already lived the workflow problem from the inside. At McKinsey, she spent six years dropping into industries she did not know, shadowing experts, and turning what they did into documentation clients could use.

The work was hands-on and slow: watch someone click through systems, take notes, capture screenshots, then translate it into a process doc or a long slide deck. It was valuable, but it did not scale.

Then the Greylock interviews gave her the same problem from the buyer’s seat. Enterprise leaders admitted they often lacked visibility into how work was actually done. Even when a process mattered, the “official” version was outdated, scattered across wikis, or missing entirely.

Put those two worlds together and the insight is straightforward: the bottleneck is not a shortage of automation tools. The bottleneck is a shortage of accurate, structured workflow documentation to automate in the first…

AI only works when you feed it clean, current information. Workflows are part of that input.

Scribe’s core behavior lines up with that bottleneck. It captures an expert’s work from their screen and turns it into a step-by-step guide with text and screenshots that can be shared and reused.

Concrete example: “How to set up a new employee in our CRM” is usually tribal knowledge. A screen-based guide makes it teachable, auditable, and easier to improve.

Choosing Documentation When Everyone Wanted Automation

Early on, Scribe leaned toward automation. That pitch fits what many investors like: efficiency, AI, fewer humans doing repetitive work.

But her interviews kept pointing to an earlier, less flashy problem. Companies were trying to automate steps they could not clearly describe, and the people who understood the work were already overloaded.

• Customer signal: leaders could not explain the real steps of critical work, so projects stalled in handoffs, training, and exceptions.
• Workarounds were the tell: teams patched the gap with screen recordings, shadowing, and one-off docs that went stale fast.
• Investor pushback was explicit: she has described investors warning that pivoting to “documentation” could kill the company.
• The operating logic: documentation is the prerequisite layer. If the workflow is wrong or missing, automation and AI efforts misfire or move risk around instead of removing work.

This is the part founders miss. “Follow investor excitement” is not the same as “follow the constraint.” She followed the constraint she kept hearing in those calls.

The Offline, ‘Janky’ Launch That Proved the Market Was Hungry

After committing to documentation-first, the team did not wait until it looked enterprise-ready. They shipped a rough MVP that was offline and clearly unfinished, then put it in the market.

They used Product Hunt as the testing surface. Not because Product Hunt is magic, but because it forces a simple question: does the core behavior land when strangers try it?

• What “janky” meant: limited functionality, rough edges, missing integrations, and offline limitations. The product showed the core action (capture a workflow, generate a guide) and skipped the rest.
• Why ship anyway: polish can hide weak demand. A rough build makes the market do more of the work, either by pulling it forward or by ignoring it.
• Signal that mattered: users sent long, unsolicited feedback. Smith has described people sending “long essays” of feedback.
• Another strong signal: people asked how they could pay. Payment intent is different from “this is cool.” It is a budget and a priority.
• Behavioral proof: internal sharing. When a tool spreads inside a company without a sales push, it is usually tied to a real day-to-day job.
• What they did not overvalue: launch-day applause, generic compliments, and surface metrics that do not change user behavior.

If you want to run a similar test, design it so you can see effort. Make users do something real, with their real workflow, in under 10 minutes. Then watch what they do next.

Scribe later reported over 5 million users and broad penetration in large companies, which suggests the early pull was not a niche anomaly. The details of the origin are covered in this PMF Show conversation with Jennifer Smith.

Mini checklist for your next “janky” launch

• Can a new user reach the “aha” moment in one sitting, without a call?
• Do you get specific complaints about missing things, or vague praise?
• Are people asking to use it with teammates, not just for themselves?
• Does anyone offer money, even in an awkward way (“Can I pay for this?”)?
• Are users doing extra work to keep using it, despite limitations?

What 1,200 Enterprise Conversations Were Really Pointing At

• Leaders could not see real workflows, only org charts. Implication: anything tied to “how work happens” becomes fragile, including compliance, quality, and automation projects.
• Critical know-how stayed trapped in experts’ heads. Implication: the company’s best practices do not scale, and the cost shows up as rework and dependency on a few people.
• Onboarding and training dragged because the steps were not documented. Implication: hiring growth is capped by training capacity, not headcount budget.
• AI and cost-cutting mandates hit a wall without process maps. Implication: leaders ask for automation, but teams cannot safely automate what they cannot describe.
• Scribe’s adoption numbers map to a widespread gap, not a corner case: over 10 million workflows documented and over 5 million users reported. Implication: “workflow capture” is a repeatable need across roles, not a one-time project.
• Reported enterprise reach (around 95% of the Fortune 500) and a reported $1.3B valuation suggest the pain was budget-worthy. Implication: documentation can be a serious enterprise category when it is tied to daily work, not policy binders.

Startup Myths Her Process Doesn’t Support

• Myth: You need the full idea before you start. Reality: she spent about three years running roughly 1,200 interviews before committing to the specific workflow documentation wedge.
• Myth: Investor excitement is the best compass. Reality: she pivoted away from the hotter automation framing toward documentation even after being told it could kill the company.
• Myth: Building is the real work, interviews are a delay. Reality: the interviews created the map, and the “janky” Product Hunt MVP tested whether the map matched real demand.

If you want a quick self-audit: look at your last two weeks. Did your calendar match the myth you say you do not believe?

Running Your Own 6-Week Version of the Scribe Discovery Play

You do not need three years or 1,200 executives to copy the mechanics. You need focus, volume that fits your stage, and disciplined note-taking.

1. Choose one target segment you will not change mid-sprint (example: IT managers at 200–2,000 person companies responsible for onboarding tools).
2. Write a 5–7 question script that forces workflow detail: “Walk me through the last time,” “Where does it break,” “What do you do when the expert is out,” “Where is this documented today?”
3. Set an interview goal of 20–50 conversations in 4–6 weeks, and block the time like it is your main job.
4. Fill the pipeline with a mix of warm intros and cold outreach. If you cannot book 5 calls a week, your segment definition is probably too vague.
5. Log every call the same day using consistent fields (role, context, workflow, pain, workaround, frequency, severity, memorable quotes).
6. Do a weekly pattern review. If you cannot summarize the top two repeated pains in one sentence each, you do not have a pattern yet.
7. Build the smallest test that targets one repeated workflow failure. It can be a basic tool, a script, or even a manual service, as long as it changes the workflow.
8. Ship a “janky” version to 5–10 people you spoke with. Make it explicit that it is early so you get honest reactions, not politeness.
9. Track validation by behavior: unsolicited detailed feedback, internal sharing to teammates, and explicit requests to pay or expand usage.

Once the Problem Is Proven, Your Job Changes

After Scribe had clear pull, the constraints moved. The company has reported over 5 million users and usage across around 95% of the Fortune 500.

Smith has said she now spends over 60% of her time on hiring and recruiting. That is a useful reminder: the “right work” changes once the market is responding.

If you are unsure where you are, look for evidence of pull first. Then change your calendar to match the stage you are actually in.