This conversation has been edited for length and clarity.
The 30-Second Splash
Omri: Is AI hype?
Amaury: No - it's a super exciting time right now.
Omri: Snowflake or Databricks?
Amaury: Whatever our customer is using. But Databricks is one of our investors, so I have to say Databricks.
Omri: Who wins the upcoming football World Cup?
Amaury: France. Allez les Bleus!
Omri: Spark or dbt?
Amaury: Let Claude Code decide what's best, depending on the use case.
Omri: Most underrated skill in data right now?
Amaury: Understanding the business case - and how the business will actually leverage the data.
Omri: And that's a wrap for the Splash. Just so you know, these answers all went on the record.
What does Nimble do, and what's your role?
Omri: In two sentences - what does Nimble do, and where do you fit in?
Amaury: Nimble is a real-time web data platform. We deploy AI-powered web search agents that browse the live internet, verify what they find, and turn it into clean, structured data your AI agents and analytics can actually use. I lead the solution consulting team, so we translate what a customer actually needs into how the product can solve it - and make sure it works in their environment before they buy.
What kind of data are companies actually trying to use?
Omri: When you talk to these customers, what data are they actually after, and what are they trying to answer?
Amaury: It varies, but almost every conversation starts with a question their internal data can't answer. A retailer wants to know what their top 20 competitors are charging right now across thousands of SKUs, and reprice before the day is out. A bank wants real-time due diligence on a counterparty - news, filings, leadership changes - pulled together in minutes instead of weeks. A CPG brand wants to track every mention of its product across reviews, social, and forums to catch a quality issue before it becomes a recall. Different industries, totally different use cases, but the same underlying challenge: the answer lives outside the company, on the open web.
So what even counts as "data" now?
Omri: Data used to mean whatever you could store in your warehouse. That's clearly not the case anymore. How do you frame what companies use today?
Amaury: The mental model we use with customers is three concentric circles. The inner circle is the warehouse - sales, CRM, product data. That's still the foundation and it doesn't go away. The middle circle is your internal unstructured data: Slack, tickets, Drive, Jira, call transcripts. The outer circle - by far the biggest - is everything outside the company: the web, third-party data, public records, social signals.
What's shifted is that outer ring. It used to be a nice-to-have - maybe you bought a dataset from a vendor twice a year. Now it's a first priority, consumed natively as part of an AI application's reasoning loop. And nobody wants something six months stale. They want to understand what's happening on the web right now to make better decisions.
Does internal data still matter, then?
Omri: Is internal data still critical, or is it stepping back to make room for all this external data?
Amaury: Internal data is still the ground truth - it's who you are. Your customer list, your product catalog, your pricing, your performance. External data is what's happening around you. You absolutely need both, and the magic happens when you join the two. The best use case isn't "let's pull a competitor dataset and look at it." It's "let's enrich every single row of our sales data with competitor promotions - by retailer and timing." Then if my sales went down, I can see the root cause instead of guessing - maybe it was a specific competitor's promotion at a specific retailer.
Why was external web data such a nightmare - and what changed?
Omri: Web data used to be a nightmare from my engineering days - endless connectors, constant data-quality issues, and the internet changing daily. How is that changing with what you're seeing at Nimble?
Amaury: It was a nightmare for two reasons. First, connection. Every website is unique - different layout, different anti-bot defenses, different JavaScript framework. Building a scraper might be a one-week project, but maintaining it was a four-hour-a-day project. Every data engineer has a story about waking up at 3am because a major retailer redesigned a product page and everything broke. Second, parsing. Even when you got the data, it came back as soup - HTML blobs, missing fields. There's a reason one of the parsers is literally called Beautiful Soup.
We built Nimble to solve both end to end. We orchestrate thousands of web search agents on top of our browser infrastructure, our own proxy network, and an optimization engine that picks the right approach for each site in real time. The agents don't just fetch a page - they run the JavaScript, navigate dynamic sites, find and validate the right fields, cross-check them, and deliver structured data straight into Databricks, Snowflake, or Azure. The customer never sees the complexity. They describe what they want and get a reliable, structured stream of web data, used the same way they'd use a table of internal data. The maintenance overhead disappears.
Does this level the playing field?
Omri: Does this let smaller organizations match what only the most advanced companies could do before? Is it democratizing?
Amaury: It's absolutely changing how this works. In the past decade, doing this meant a team of 20 engineers - so a massive tech company or a hedge fund could, but a mid-market retailer? Forget it. They'd buy a static dataset once a quarter and call it competitive intelligence. Now mid-sized companies are standing up agentic workflows that consume live web data in production and reprice daily or even hourly. They're not building proxy infrastructure, maintaining headless browsers, or parsing HTML. They just describe what they need in natural language - through APIs, or through Claude Code, Cursor, Codex - and get structured data back. That's a massive democratization. The moat used to be the infrastructure. Now the moat is what you actually do with the data.
Omri: So the way I think about my data estate becomes the moat - not the technical capability. It's about getting analysts to think about what questions they want to ask.
Amaury: Absolutely.
Has real-time finally stopped being a buzzword?
Omri: Real-time external data was a buzzword for years. Has that finally changed?
Amaury: Absolutely. Real-time isn't a nice-to-have anymore - the value compounds. Every minute of freshness translates directly into better decisions and faster reactions, especially in retail and CPG. If something was on promotion yesterday or last week, who cares? It's already too late. We work with a very large pizza chain that wants to know, for nearly every zip code in the US, what competitors are offering - large chains and mom-and-pop shops alike: what menus, what promotions. And the offering looks different at 7am, noon, and 9pm. So it has to be real-time, multiple times a day, across thousands of restaurants, so they can reprice and make the best decisions.
How does this change the pipelines themselves?
Omri: With a lot of internal and external data, how does pipeline-building change? Is it still store-everything-then-query, or more ad hoc?
Amaury: We see both, and it depends on the use case. A retailer might say, "I want entire categories from Walmart, Target, Amazon every morning, or multiple times a day, to understand the full competitive landscape." There you fetch it fresh, store it, and build trends over time to see what's emerging. Other cases are ad hoc - the World Cup might be one, where you just want to know which companies are running banners or game-specific offers, fetched on demand. We also work with management consulting firms that break a big question down - say, the market size for a new type 2 diabetes drug - into smaller ones, use Nimble to search the web for each, and assemble the answer the way a consultant would. Usually it's a combination of both.
What does this look like in practice?
Omri: Give me a concrete example - and did the customer have to rebuild everything to handle the influx of data?
Amaury: One of our customers is one of the biggest beverage companies in the world, and what they care about is where to send their sales team. So on a daily basis we capture the menu of every single restaurant in Japan to understand which sodas and beverages are being sold, and at what price. That tells the sales team exactly which door to knock on - where a competitor's product is on the menu, at what price - so they can optimize their efficiency.
And no, the implementation was fairly straightforward. The tables were already there; they were already capturing this data, just manually or at very low scale. What changed was the scale, the volume, the frequency, and ultimately the sample size they have to make business decisions. There was a change-management process on top - they built a dashboard with color-coded regions showing where the high-value, lowest-hanging-fruit areas are - but the data model itself didn't have to change.
Where is the data stack going in the next three years?
Omri: Three years out - how much is internal versus external, batch versus ad hoc, and humans versus agents consuming it?
Amaury: Given how fast AI moves, predicting three months out is tough, but I'll try for three years. First, the line between internal and external data starts to disappear for the end consumer. Whether a row came from your CRM or from a website rendered ten seconds ago, the analyst - and more often the agent - won't care, and shouldn't have to.
Second, most data will be consumed by machines, not humans. Dashboards are still the primary interface today, but in three years agents will be the dominant consumers of data. That changes everything about how you design the stack - token efficiency, structured output, latency, and schema clarity matter much more than dashboard aesthetics.
Third, real-time stops being a category and becomes a default property. Saying "real-time data platform" in 2029 will sound like saying "color TV." Of course it's in color. Of course it's real-time.
What should a data leader do tomorrow?
Omri: A data leader is listening right now. What do they do tomorrow to start working with external data?
Amaury: Pick one decision your business is making blind right now because you don't have external data - pricing, competitive intelligence, supplier risk, candidate research. Something specific. Not a strategy, just a decision you're making blind.
Then don't boil the ocean. Get web data flowing into your existing environment, joined to internal data, in front of one team - a two-week project, not a two-quarter one. And measure what changed. Are you making faster decisions? Better outcomes? More automation? If you can't point to what changed, you probably picked the wrong use case - so go back to step one. And by the way, go to nimbleway.com, create an account, and start using it - through a direct API call or plugged into Claude Code, Cursor, Codex, whatever you use - and get data very quickly.
Closing takeaways
Omri: A few takeaways I'm leaving with:
- Real-time data is no longer a category - it's an expectation. Soon we'll stop saying "we have real-time data" the same way we stopped saying "color TV." It's just assumed.
- External data is becoming a first-class citizen in the stack. We'll stop asking where the data came from and simply expect both internal and external data to be there, joined together, driving decisions.
- Start from the question, then bring in the right data. Stop asking "what data can I pull in and what can I do with it?" All the data is going to be available - so decide what you want to answer first, then go get exactly what answers it.
If you enjoyed this conversation, follow the Data Splash podcast. We have a lot more coming. Until next time - keep your data flowing.
For the full episode:
Youtube: https://youtu.be/Y8V69YrzHIY?si=bSo6zHC9bVPXG3Ep
Spotify: https://open.spotify.com/episode/2ydr0OhstutmeMxSChIaxn?si=85syjwNURGKkNQdyzwtfIA
To follow the Data Splash:
Youtube: https://www.youtube.com/playlist?list=PLJboMtNQp-MtIK1v7fn6eA35Ylh2T85IN
Spotify: https://open.spotify.com/show/4MD8LSNbpzW46qeq3qptpr
