Lumi AI nets $3.7m to expand conversational analytics platform

[Disclosure: AgFunderNews’ parent company AgFunder is an investor in Lumi AI.]

• Lumi AI—a startup enabling users without coding skills to gain actionable insights from complex data in company ERP systems using plain language prompts—has raised a $3.7 million seed round led by AgFunder.
• The round, which was supported by Forum Ventures, Abu Dhabi sovereign wealth fund ADQ, and strategic angel investors, will help the Toronto-based firm expand its platform to support a growing roster of customers including a leading US supermarket chain, a top meal-kit provider, a leading consulting firm, the largest retail operator in the Middle East, and mid-sized CPG companies.

Founded in 2023 by Ibrahim Ashqar and Tudor Boiangiu, Lumi AI has built an enterprise data analytics platform with an emphasis on supply chain data enabling users to interrogate enterprise resource planning (ERP) systems via a natural language interface.

The interface, which integrates with clients’ supply chain software, operates as an intelligent assistant, helping users retrieve and interpret data, generate charts and extract tailored insights that previously would have taken days to pull together, says Ashqar.

“We founded Lumi AI to break down the barriers between businesses and their data. With this funding, we’re accelerating our mission to help companies democratize access to intelligence and transform decision-making at scale.”

“Lumi AI is transforming the way businesses operate by making analytics accessible to everyone. I haven’t seen that before – such incredible pull in demand across retail, grocery, manufacturing – any industry with complex supply chains.” Rob Leclerc PhD, managing partner, AgFunder

Eliminating data bottlenecks

Typically, if business users want to extract insights from ERP systems today, they will file a request with an in-house data team or someone in the company that can code in SQL or Python to generate a custom report, Ashqar tells AgFunderNews.

“Whenever you have a situation like that, it creates a bottleneck,” says Ashqar. “The data team is overloaded with requests and business teams are frustrated because they’re overly dependent on the data team.”

And in the meantime, he says, “Lots of opportunity remains hidden in the data and you’re leaving money on the table.”

And this is the basic premise behind Lumi AI, which is “all about illumination… shedding light on issues within your data sets,” says Ashqar, who met his cofounder Tudor Boiangiu while working at Deloitte’s AI practice.

“Large language models came on the scene in late 2022 and they demonstrated an incremental ability to code, which opened up a world of possibilities. So I mocked up a proof of concept, showed it to Tudor, and we ended up building an AI data analyst offering agentic workflows for analytics, empowering people to get the most from their data just by asking questions in plain language.”

Use cases

Initially, Lumi AI is targeting companies with supply chain data models that use platforms such as SAP, Oracle, or Microsoft Dynamics to run their operations, he says.

Typically these are companies with a lot of inventory data, sales history, procurement history, customers, vendors, bills of materials, shipments, and so on, he adds. “We’ve spent a lot of time in this space so we know the data quite well, so it made sense to start here as opposed to saying we can do everything for everyone.”

In the case of a supermarket chain, for example, he says, on-shelf-availability is a critical metric. “So your issue might be, the distribution centers are full of inventory, so why are we seeing out of stocks?”

Often it’s because some stores are not following inventory protocols properly, he claims. In order to identify where the problem lies, historically, account management and retail ops teams would have to go through dashboards, manually extract data into spreadsheets, perhaps ask an analyst to bring in new data, conduct an analysis, “and after about a week or so, they will conclude, ‘We need to talk to these five stores.’”

With Lumi AI, he says, these teams can simply ask which stores are not following the protocols in plain language, quickly identify the offending stores, and take remedial action.

Another key use case is identifying discrepancies between what is ordered and what is shipped, he says. Big retailers collect this kind of data at a very granular level, but there’s so much of it, it’s “almost unusable if you want a dashboard view,” he says. “So to see what’s going on, they have to aggregate it to a higher level, which means they lose visibility and can’t see what’s happening at the item/store/day level.”

Lumi AI, however, enables users to “drill down to the most granular level so you can see what’s actually driving a dip or an anomaly or an outlier.”

Tacking ‘dashboard anarchy’

When it comes to persuading potential customers of the value of Lumi’s offering, he says, it depends on the maturity of the client. “Mid-sized clients typically don’t have a data team, or it’s really small, so we’re giving them ability to get insights they didn’t have access to before.”

For such clients, Lumi can also create pre-set prompts that ask questions such as, ‘Show me items where the minimum order quantities from our vendors far exceed the average consumption for these raw materials.’ This could be a good starting point for negotiations with vendors that are “making you buy way more inventory than you need, and it’s eating up your cash flow,” says Ashqar.

For enterprise clients such as big retailers, he says, Lumi is offering a cure for “dashboard anarchy,”  a condition characterized by “having hundreds of dashboards, 95% of which are useless.”

As for barriers to progress, he says, on rare occasions, potential clients might have a system that Lumi AI can’t connect to. “Shopify gatekeeps its data, for example, so Shopify customers can’t actually directly query it. First you have to extract the data via API, migrate it to your own data warehouse, and then you can query it. So some potential customers haven’t yet got their data to a place where we can query it.”

Addressing data security concerns

Importantly, says Ashqar, companies deploying Lumi AI do not have to worry about data security. “From day one, we designed Lumi in a very specific way that ensures that the raw data is only ever processed within the client’s infrastructure. Large enterprises love that because we’re not telling them, ‘Share your data with us and then we’ll give you insight.’”

As for onboarding, he says, “You can’t just connect AI to your data and expect magic. It needs to understand the context of your data and your business for it to work properly. But we’re talking days and weeks, not months. With the biggest retailer we are working with, we got them up and running within a week.”

Agentic workflows

Moving forward, he says, “We need to continue innovating because the field is moving so quickly.  For example, a lot of the ideas that Tudor had at the outset were pioneering, like the concept of agentic workflows [whereby the AI makes decisions and takes actions autonomously based on predefined instructions and goals], a term that did not exist at the start of this hype, and is now almost mainstream.”

Longer-term, he says, Lumi’s real value will come not just from answering users’ questions, but from creating a system that can start asking itself questions and then flag anomalies.

“Today, Lumi AI is a senior data analyst. The next evolution is it becomes an analytics manager that can proactively surface insights.”