The Startup That’s Using AI To Take On The Grocery Giants

“In a world where everything can be delivered to my doorstep, why am I wandering around a Safeway at nine o’clock at night trying to buy milk — again?”

It is a thought that has occurred to most parents in some form or another during the first few years of their children’s lives. Maybe they substituted “yogurt” or “bananas” for milk; perhaps they thought “Walmart” or “Whole Foods” instead of Safeway, but no matter the specifics, the sheer number of pajama-clad nighttime trips to a grocery store drives every parent to the same existential crisis that hit Pradeep Elankumaran shortly after his daughter turned two and developed a major milk habit.

“A lot of milk was being drunk, and suddenly I am at the store three or four times a week buying the same things over and over,” he said. “And somewhere around the 30th iteration of this experience, I started wondering about how I could automate this. Because I am an engineer, and we don’t want to do the same thing over and over.”

From AmazonFresh to Instacart, Elankumaran said he and his wife had tried just about everything and everyone they could think of to keep their daughter in milk and themselves off the grocery store graveyard shift.

“But something always happened; it never quite worked right. Something always came up that muffed up the process,” Elankumaran said.

So, he posted his problem to a local message board to see if there were others in his community struggling to keep a simple supply of staples coming to their doorstep. He knew there was a business in the making when 200 people posted the exact same problem within 48 hours — all wishing someone, anyone, could make grocery delivery work for them.

With that, the idea for Farmstead was born, Elankumaran told Karen Webster in the latest The Matchmaker Is In.


An AI Company with a Small Footprint and a Big Ambition

Today, Farmstead is fairly small — its delivery footprint is limited to San Francisco and the Bay Area, and it doesn’t carry the 40,000 SKUs that most grocery stores do. In fact, Farmstead offers only about 1,000.

On purpose.

Elankumaran noted that Farmstead is “a very opinionated” company in this regard and does a lot of curation offering “category coverage as opposed to SKU coverage.” It’s quality, not quantity, that matters.

That gives Farmstead a sourcing advantage, too. Elankumaran said that grocery stores have their own proprietary relationships with producers and suppliers, given the large quantities of product they have to buy. Instead, Farmstead works through the same supply lines that local restaurants do when they are sourcing their perishable goods (and with national and local brands for non-perishables), hence they skew incredibly local and are able to deliver extremely fresh food to their customers’ doors.

They deliver that food very fast. Farmstead customers can choose to have fresh food at their doorstep within 60 minutes, by the end of the day — or they can set a weekly recurring order to show up on their day of choice.

Farmstead started about a year ago with its scheduled weekly delivery. That enabled them to keep costs down by only ordering just enough supply to fulfill those orders, so they didn’t lose money on unpurchased product that would go bad days later.

But that also presented a friction point for buyers, Elankumaran discovered, who felt that pre-ordering groceries was an anxiety-producing change in their behavior.

“What we have seen in the 16,000 deliveries we’ve made in the last year is that we have lots of demand for on-demand,” Elankumaran said.

Farmstead has set itself apart by allowing customers to obtain on-demand orders from the same place as their standing, weekly orders. According to Elankumaran, this changes the dynamics of grocery shopping, as it can become a household utility service around which customers can manage their workflow.


Unchaining the Power of AI

Farmstead identifies itself as an artificial intelligence-powered tech company that happens to sell groceries — because, as Elankumaran said, their proprietary AI platform allows them to do things more efficiently than other firms.

Because Farmstead offers same-day (and same-hour, in some markets) grocery delivery, they must estimate how many goods to keep on hand in their micro-hubs that supply deliveries. The AI, as Pradeep Elankumaran noted, is sophisticated enough to stack up not just an individual buyer’s information, but whole regional buying patterns over time — and can make very specific predictions about the types and quantities of inventory in each area Farmstead serves.

That smart software, he pointed out, also has the all-important job of optimizing and perfecting delivery routes for drivers, ensuring that deliveries are made in the most efficient way possible both in terms of time and miles driven.

Although Farmstead now serves San Francisco, it started out in the Bay Area tackling the problem most urban innovators want nothing to do with: the low-density suburbs.

“Suburban areas are where a lot of logistics work comes into play, particularly around optimizing routes, algorithmically,” said Elankumaran. “And that is a very detailed process — things like terrain grade and whether the area is flat or hilly end up being things you spend a lot of time thinking about and building into the route planning,”

This, he noted, is why Farmstead has been so circumspect about opening new markets, choosing instead to go zip code by zip code as it establishes an audience and supply lines around the micro-hubs that house its supply, servicing areas in a 50-square-mile radius.

“This particular problem is why online grocery endeavors before Farmstead have failed,” said Elankumaran. “This is all very hard to do in real time, on demand and at the right density. We have a team that has spent a lot of time thinking about that question — both for us, and for other firms before they came to us.”

But done right, the flywheel gets started when one customer tells another that the service is awesome and brings them along, which gives the hubs more demand to aggregate, which eventually (with enough new additions) improves their margins enough so they can lower their prices. As that tends to stimulate more use, more demand also means routes become more dense and thus efficient to drive, which further improves margins and stabilizes costs.

“What we see in many markets is eventually we have best-in-class prices to go along with our best-in-class, highly fresh produce,” noted Elankumaran.

While this model is only online in the Bay Area today, Elankumaran says the model — and building density and demand in micro-chunks across regions — is a scalable model that will eventually be available nationwide, or even worldwide.

It’s a big goal, but, as Elankumaran learned during his first 48 hours thinking about this business when he posted on a message board a year ago, there is a lot of demand out there going unmet.

And innovators abhor a vacuum.