LightSource is a procurement AI company that helps businesses collaborate with suppliers to launch products faster and at structurally lower prices. The company focuses on direct procurement, bringing engineering teams, procurement organizations, and suppliers onto a common platform supported by AI-native workflows. Pulse 2.0 interviewed LightSource CEO and Co-Founder Spencer Penn to learn more.
Spencer Penn’s Background
When asked about his background, Penn shared:
I’m Spencer, CEO and co-founder of LightSource, a procurement AI company. We help businesses collaborate with their suppliers to launch products faster and at structurally lower prices.
I’m a New Yorker by birth, Manhattan specifically, but I moved to the Bay Area 11 years ago to join Tesla back when it was still a boutique car company. When I told friends and family I was going, the consensus was that I’d lost it: “EVs will never be anyone’s main car. It’s a tech toy, an iPad with wheels on it.” That aged well.
I went anyway because I wanted to be part of a next-generation hardware business with a bona fide environmental impact. What I walked into was chaos, the good kind. I helped lead the development and launch of the Model 3, Tesla’s first mass-market EV, and got a front-row seat to every problem you can have building a car at scale.
Plenty of those problems were interesting. But the one that wouldn’t leave me alone was procurement. We were managing $30B of spend through spreadsheets and email. It just drove me nuts. A year or two after leaving Tesla, I couldn’t stop asking myself, what would it look like to actually try to solve this?
How LightSource Started
When asked how the idea for the company came together, Penn explained:
The idea crystallized at Harvard Business School. I was in a class where the final project was building a startup concept end-to-end, and I used it as an excuse to pull on the thread.
What surprised me wasn’t that Tesla had this problem. I knew that. It was that everyone had it. I’d talk to procurement leaders across the Fortune 500 and hear the same story, almost word for word: Excel and email, Excel and email.
By comparison, the sales side of those same companies was running on Salesforce, a customer 360, pipeline analytics, the whole stack. Procurement and Sales are two sides of the same transaction. Sales sells to Procurement; Procurement buys from Sales.
So it only makes sense that there should be a “Salesforce in reverse,” a system that helps procurement teams find vendors, buy from them more effectively, and run a real strategic sourcing motion. We’re building Salesforce in reverse, for the trillions of dollars that flow the other way.
The problem statement is just one piece. Three other things turned the dream into a real company: (1) finding a great business partner in my co-founder and CTO, Idan, (2) getting some early customer traction, and (3) investor interest.
That combination gave us the conviction to leave our cushy jobs at Waymo and Google Research and dive in headfirst.
Favorite Memory
When asked about his favorite memory working for the company so far, Penn recalled:
It’s hard to pick just one, but the moment that comes to mind is closing our first round of funding.
It was during COVID, so the whole thing happened over the phone. Idan and I were both live in a Google Sheet, modeling the deal in real time as we talked to our investors. We almost had numbers we could align on.
While I was on the call, Idan and I were typing actual messages to each other in the cells. I’d write a question in one cell, he’d respond in the cell below. There’s still a hilarious 30-cell stretch in that sheet that reads like a text thread.
At a certain point, I asked Idan for his blessing to move forward. He typed back: “close it dude.” And we had a deal.
That moment has stuck with me because it captured something true about how Idan and I work together: fast, in the same document, no ceremony. Same way we run the company today.
Core Products And Features
When asked about LightSource’s core products and features, Penn detailed:
LightSource is an AI company focused on direct procurement, not the office odds and ends, but the materials and components that manufacturing businesses put into their products. Think of the steel, the wire harnesses, the battery cells, the precision-machined parts.
Today, there are major silos between engineering teams, procurement, and suppliers. Processes are highly manual, prone to error, and lead to worse outcomes for everyone, including the end consumer who pays more and waits longer for the product.
With LightSource, we bring all three of those groups onto a single platform. Our core features include BOM management that lives alongside the engineering source of truth, a flexible sourcing and RFQ engine, bid analytics that help buyers compare apples to apples across complex multi-line quotes, supplier discovery and benchmarking, and an AI-native supplier-collaboration surface where engineers, buyers, and suppliers can actually work together.
The whole stack is built so the AI handles the manual coordination, pulling specs, chasing quotes, normalizing supplier responses, and the humans focus on the judgment calls that actually move the program.
The result: engineers, buyers, and suppliers working from the same source of truth, and customers launching products faster at better prices.
Changing Procurement Software Expectations
When asked about challenges in the sector and how LightSource has addressed them, Penn noted:
Plenty. The biggest one is that customers have been burned by procurement software before. Many of them have lived through 5-to-10-year SAP transformations or Ariba and Coupa rollouts, and the experience was almost universally bad: insanely expensive, painfully slow to implement, and at the end of it, the software is so clunky that it makes their jobs harder, not easier.
Worst of all, those tools were built for indirect spend, buying stuff, not for the strategic direct materials work that’s actually core to the business. Even when they’re implemented well, it’s a square peg in a round hole.
That’s a big reason buyers fall back to email and Excel. It’s not automated, there’s no data capture, but at least it doesn’t fight you.
So a big preconception we have to disabuse people of, fast, is how painful procurement software has to be. Honestly, one demo usually does it.
When folks see LightSource autonomously ingest a BOM and launch a sourcing event in under five minutes, you can watch their faces change. They know it’s a new dawn.
Technology Evolution
When asked how LightSource’s technology has evolved since launching, Penn explained:
The AI space is moving faster than people can publish blog posts about it. We were lucky in some ways. We put our chips on an AI-native architecture early. For reference, we registered the domain LightSource.ai in July 2020, almost six years ago, well before the current Gen AI craze.
When we started the company, transformer models were a recent invention and were good at language tasks. (That’s literally what LLM stands for, Large Language Model.) The infrastructure that’s grown up around those models in the years since is what’s allowed people to do a much broader range of work with them.
We’ve stayed AI maximalists in every sense, not just in the product features we ship to customers, but in how we work internally as a company.
Not only do our technical teams use agentic systems to assist with coding and testing, but our deployment teams clean, manipulate, and understand customer data using AI tools that compress migration timelines from years to days.
The biggest shift in the last year is that models have gotten good enough to operate inside dynamic environments, not just generate outputs, but take action on behalf of the user, iteratively. That’s what “agentic” means.
We’re again leading the pack on bringing that next level of autonomy to procurement.
Major Company Milestones
When asked about some of LightSource’s most significant milestones, Penn said:
Honestly, every week feels like we’re re-founding the company.
The first and most important milestone was meeting Idan. Without him there’s no company. The second was hiring our first engineer. It still amazes me that someone chose to join us when we had no money, no product, and no customers. That’s brave.
The next big one was landing our first customers. I have a lot of appreciation for early customers. They’re making a bet on unproven technology, and in most companies you don’t get rewarded for taking a big risk that pays off, but you can absolutely get burned if it doesn’t. Sticking your neck out for a startup is a real act of faith.
Raising institutional funding took LightSource from a project to a functional company. Fundraising was never a goal in itself. What excited me was the ability to actually run the experiment at full scale.
We have big ambitions, and pulling them off takes the right mix of energy, team, customer trust, and backers.
Automotive Customer Success
When asked to share a specific customer success story, Penn highlighted:
Sure. One of our customers is a Big Three auto manufacturer that developed one of their most important programs in recent years on our platform.
But here’s the thing. They only brought LightSource in halfway through the development cycle. At first I was a little peeved, because I wanted the entire BOM running through us. But it turned into a happy accident: a built-in A/B test.
Same vehicle, same company, same buyers. We could compare side-by-side the parts and systems sourced through LightSource against the ones that weren’t.
The results landed in two places: speed and cost.
On speed, they accelerated their sourcing cycle, the time from engineering design to supplier selection, by 25%. Repeat that across hundreds of RFQs and thousands of parts, and you’re accelerating the whole vehicle program by a meaningful amount.
On cost, the impact wasn’t savings per se. It was cost creep avoidance. Programs always start with optimistic cost profiles. You get great prices when suppliers are competing to win the business. The minute you pick one, you have no leverage, and every design change after that turns into a cost-up. That’s cost creep. Death by a thousand paper cuts.
We found cost creep was reduced by 47% on the components sourced through LightSource versus the ones that weren’t.
Funding And Revenue Growth
When asked about funding and revenue metrics, Penn shared:
To some degree, yes. We’ve been on a tear. At our company holiday party last December, we were 18 people. This past December we were 60.
I did this thing during my speech where I asked folks to raise their hand if they’d started in the last week, last month, last quarter, last year, and it was amazing to watch nearly every hand in the room go up.
On revenue, I’m proud to say we’ve posted triple-digit growth four years running.
On funding, we announced a $33M Series A in early 2025 at a $130M valuation, led by Lightspeed and Bain Capital, two all-star investment teams.
We’re proud not just of the capital, but of the partnership we’ve had with our board and the operating partners around the table.
Market Opportunity
When discussing the total addressable market LightSource is pursuing, Penn explained:
The funny thing is that the procurement software market today isn’t that big. When we go to DPW or ISM, the largest procure-tech conferences, we notice they’ve gained real momentum but still cap at 1,000 to 2,000 attendees.
By comparison, Dreamforce alone is something like 25,000 people, has been headlined by the Red Hot Chili Peppers, and basically takes over the city of San Francisco. You don’t see that in procurement yet, even though, mathematically, the dollar value that sales teams sell is equivalent to what procurement teams buy.
So the TAM for procurement software is around $9B globally, and broader supply chain software is around $25B. But the World Bank pegs global manufacturing value-added, the GDP-equivalent for manufacturing, at $16.8T.
If you think first-principles, procurement is the largest flow of money in almost any business, and the people running it are the least well-equipped with modern software.
So I think less about the $9B procure-tech market today and more about the $17T of purchasing that direct materials teams manage. It’s a bit like looking at the world in 1980 and asking, what’s the TAM for CRM? It’s a lot bigger now.
Competitive Differentiation
When asked what differentiates LightSource from its competition, Penn identified three areas:
Three things:
- We focus on direct materials. There are broadly two flavors of procurement, direct materials (the things a company buys to put into its products) and indirect materials (anything a company buys to consume itself).
When I was at Tesla, we bought $30B of car parts (direct) and also office chairs, toilet paper, soap (indirect). There are hundreds of tools that help companies manage indirect or tail spend, P2P systems, Amazon Business, purchase cards, even employee reimbursement counts.
In the direct materials space, there’s almost nothing that solves the actual pain points buyers face. That’s the gap we’re built for.
- We’re AI-native. We’re built from the ground up to incorporate and interoperate with AI and agentic systems. Every feature in LightSource either draws from or contributes to a common harness, so the value compounds. The more you use the platform, the smarter it gets for you.
And we treat the system as a proactive partner that helps get your work done, not just a record-keeping tool.
- We build for collaboration between engineering, procurement, and suppliers. Most procurement tools are built for, well, procurement teams. But in direct materials, especially during new product introduction (NPI), there’s a huge amount of design iteration happening across engineering, the buyer, and the supplier.
A drawing changes, a tolerance gets tightened, a supplier flags a manufacturability issue, and that ripples through cost, timeline, and the next round of quoting.
In traditional setups, all of that happens over email and dropped files, and critical context gets lost between teams. LightSource brings all three parties into the same workflow, so engineering changes, supplier quotes, and buyer decisions stay connected to the same source of truth.
That’s how you actually run direct procurement, not as a series of handoffs, but as a continuous, multi-party motion.
Future Goals
When discussing LightSource’s future goals, Penn said:
Our number-one goal is to help our customers deliver world-changing products to market faster and at more competitive prices. We help companies win on speed.
To do that well, we have an ambitious roadmap that’s right at the edge of feasible, and we have to recruit and retain the best team in the world to pull it off.
And finally, if we’re successful in our mission, I’d like to see us IPO, not as an end in itself, but as a way to bring a much larger group of people around the table in support, and in literal ownership, of what we’re building.
AI-Native Versus AI-Flavored
When invited to discuss another topic, Penn concluded:
One thing I think is worth saying: there’s a lot of noise right now about agentic AI in the enterprise, and not enough signal. Every software company on earth is slapping “AI-powered” on their landing page. Take it seriously, not literally.
Here’s what I’d watch instead. The companies that are going to actually deliver agentic value to enterprise customers are the ones who’ve rebuilt their products, and frankly their internal engineering culture, around the assumption that AI is a first-class operator, not a feature.
That means rethinking your data model so an agent can actually traverse it, rethinking your UX so a human can supervise an agent’s work, and rethinking how your team writes software so that AI is part of the loop, not bolted on after the fact.
We’ve been operating that way at LightSource since well before it was the consensus. Our engineers run agentic coding workflows at full agency. Our deployment team uses AI to compress what used to be multi-year data migrations into days.
And the product itself is built so that every feature feeds a common harness. The more it’s used, the smarter it gets.
The next few years are going to sort the AI-native companies from the AI-flavored ones pretty quickly. I’d just encourage readers, whether you’re buying software or building it, to ask the harder question: is this product actually built around AI, or is AI sprinkled on top? It usually doesn’t take long to tell.
