Flatiron Health cut its feedback cycle 75% with Airtable AI
How a small product operations team turned scattered sheets, documents, and tribal knowledge into a product operating system in Airtable and AI that reads customer feedback at a scale.
75%
less time to process customer feedback — from four weeks to one
4 → 1
people needed to run quarterly feedback synthesis
30 min
to build an account plan that used to take five to six hours
When a customer submits a feature request to Flatiron Health, Airtable AI helps the product team begin analyzing it immediately. It identifies the underlying user problem, searches years of feedback for similar requests, explains why each result may be relevant and suggests follow up questions for the account manager to explore with the customer.
By the time a product manager reviews the request, much of the initial analysis is already complete.
The workflow is part of a connected product operating system built in Airtable by Kassim Alani, who leads product operations at Flatiron Health, and his seven-person team. Flatiron Health is a healthtech company dedicated to improving cancer care and advancing research using real-world data. Its product organization manages more than 40 team roadmaps.
Today, information about what is shipping and when, how R&D investments align with company strategy, what customers need, and how the market is evolving lives within one connected system now.
With Airtable, Flatiron Health has:
- Reduced customer feedback processing from four weeks and four people to one week and one person
- Built a market and customer intelligence platform it would never have taken on before with a 1+ year estimated build time
- Reduced account planning from five hours to thirty minutes
- Created analysis the team had abandoned for lack of capacity
- Given R&D, finance, and the executive team one set of numbers to plan from
Sheets, documents, and people's heads
As Flatiron's product organization grew, so did the number of teams, launches, and initiatives underway. Individual groups tracked their work using the tools and processes that worked best for them, making it difficult to maintain a complete view across the organization. Kassim's team is responsible for knowing and understanding what every product team at Flatiron is building, but before Airtable, that information did not live in one connected place.
"Organizations run in a lot of different operating models, where information is often scattered about — sheets, documents, people's heads. That becomes really complex as your organization grows or scales: keeping coherence from the top all the way down to the bottom, where the work is happening."
That fragmentation created a recurring challenge. When finance asked how much R&D investment had gone toward discretionary versus non-discretionary work, answering the question meant opening six spreadsheets from six product groups and reconciling them by hand. The process could take up to a week or longer.
Without fast, accessible data some decisions relied more heavily on instinct. Dates could slip, risks were harder to identify, and the people holding context spent significant time transferring information between teams instead of solving problems.
The immediate challenge that led Kassim to look for a different tool, however, was much small and more constant than any of that.
"As someone who is really upset about our myriad of Google Sheets, where the same object was just replicated over and over again, I knew there were tools that existed."
Airtable was already cleared for use inside Flatiron. He could start building instead of opening a procurement cycle.
From launch tracker to product operating system
Kassim's team first began by creating a base in Airtable that organized initiatives live in one table, teams in another, product lines and objectives in their own, with linked record fields connecting them. A team stops being a word retyped in six files and becomes one record every initiative points at.
The team initially used the base to organize a growing number of feature launches. But each launch was already linked to its team, product line, and objective, so each new question the business asked could be answered without building anything new. Market insights, customer feedback, and competitive intelligence followed, each a new Airtable table hanging off the same records.
"We started locally on 'what are we working on and when will it ship,' all the way to 'what are the inputs to what we build.”
The team then applied the system to quarterly R&D reporting. Working with finance and R&D leaders, the team agreed on a small shared taxonomy and applied it to the roadmap records already in the base. Each initiative maps to exactly one investment theme, one objective, and one product line, so the percentages total a clean hundred.
What makes it stick is how little anyone has to do. A product manager submits an initiative through an Airtable form and maps it once. Each quarter, they enter a single number: the engineering weeks they expect to spend. Directors open a filtered Airtable view for about half an hour to catch missing entries and confirm mappings. Executives open an interface built for one question, where the answer is already clean.
Three roles, use three tailored Airtable surfaces, all drawing from the same underlying records. The number the board sees is the number the product manager entered, and the same records now shape the case the team brings to Flatiron's executive team for strategic planning. Conversations that used to run on perception run on evidence instead, and as Kassim puts it, the system “creates a place to use shared language.”
"Limit what you ask people to enter, audit what they entered, and let the structure carry the weight."
Using AI to analyze feedback, roadmaps, and the market
With that foundation in place, Kassim's team introduced Airtable's AI layer, and work the team had written off as too expensive became feasible again.
Kassim describes customer insight as a pyramid. Feedback from major customers sat well covered at the top, and classic user research held the middle. The bottom layer was everyday feedback arriving in high volume and low signal, and nobody had the hours to read it.
"Basically, what we've done is filled the gap of insight because of Airtable AI."
Hundreds of feedback items a month, one person to process them
Flatiron had previously tagged customer feedback manually but discontinued the process as volumes increased. Now, Airtable AI field agents read every one against the same prompt:
- Name the core problem.
- Pull out any feature request.
- Find similar feedback from the past 90 days.
- Suggest the questions worth asking.
Because those answers land as structured Airtable fields rather than in one person's chat window, anyone can slice them by workflow, customer, or quarter. A quarterly process that previously took four people four weeks now takes one person one week.
"Reviewing hundreds of pieces of feedback each month now becomes feasible, as opposed to feeling like a trade-off against other priorities."
Giving account managers a stronger starting point
The team built Airtable AI agents that define the user problem behind each feature request, generate follow-up questions for the account manager to ask, and help them prompt Omni, Airtable's AI assistant, to surface similar past requests.
The same field agent approach handles roadmap routing where agents read each roadmap document and extract team names and dependencies, matching them back to team records so the right roadmap reaches the right stakeholder rather than leaving someone to guess which plan applies to them.
"We've built agents that can take a doc URL, pull out the signal, and turn it into something actionable, like mapping team relationships and dependencies in a way that's actually visible, or automatically routing the right roadmaps to the right stakeholders so people aren't staring at 50+ team plans wondering, 'Which one applies to me?'"
Market intelligence built in a month, not a year
For Flatiron's life sciences business, the team is connecting datasets that were never designed to meet: thousands of trials, drugs, and events alongside opportunity data from Salesforce. Deep Match, an Airtable field agent that finds and links the most relevant records across tables, does the joining without the custom scripting it would otherwise require.
A data insights engineer had scoped that work at a year of effort. Kassim's non-technical product ops team had an MVP running in about a month and a scaled version in three. In his telling, this was never a project that got faster. It was a project that would not have happened without Airtable.
"It makes problems that were otherwise inconceivable to solve very solvable."
Flatiron operates within the heavily regulated healthcare industry, so the system deliberately holds only product operations data: roadmaps, team structure, investment allocation. Nothing clinical goes near it. And the division of labor stays deliberate, with agents reading, matching, and drafting while product managers, account managers, and directors decide.
"Not to replace their function, but to supercharge their work."
The rollout that failed and the interface that fixed it
The first rollout to a large group skipped Airtable's interfaces entirely and had everyone edit tables directly. "It was such a mess," Kassim says. The second attempt gave product managers an Airtable interface that pulled their monthly asks into one place, and the reaction flipped to "Oh wow, this is so much better."
"Sometimes you need to launch poor MVPs, even if you can launch a better version. What people appreciate is incremental progress and listening to feedback."
The rest is practical. Ask for as few actions as possible, and build an Airtable form rather than requesting direct entry. Learn which views people actually like: "People hate list views at Flatiron. They love grid views. Build a grid view," Kassim says. And when nobody uses what you built, take it apart without sentiment.
Finance became a partner and teams stopped guessing
The reconciliation week is gone, and the relationship around the number changed with it.
"Finance has become our partner instead of an interrogator. We can surface insights proactively before we're even asked."
The change Kassim watches for most closely doesn't appear on any dashboard. It shows up in what product teams reach for when they want to build something new. "When product teams propose new things they want to build, they are referencing the tools and resources we built," Kassim says. "We can have an opinion before we even go to the user or customer — which unlocks the value of synchronous time by having done our homework."
The organization has also gotten steadier through change. The context now lives in Airtable rather than in one person's head.
What comes next is a shift in kind. "What we're trying to do is basically go from building aggregation tools to insight tools," Kassim says.
The team is scaling the market intelligence platform and mapping the relationships between teams to inform how the organization designs itself. The goal for the year is that ten percent of the company regularly asks Omni, Airtable's AI assistant, questions about customers, R&D work, and the market, and trusts every answer.
Colleagues call him "the Airtable guy." He puts the job differently.
"The real superpower is using systems to create coherence — taking scattered information and turning it into a shared, structured way of working that helps teams focus on outcomes."