Supply chain / Grocery AI · Deep dive

Afresh Technologies

AI ordering engine for fresh and perishable grocery categories — flagship Albertsons rollout across 2,200+ US stores, ~$148M raised across seed, Series A ($12M, March 2020) and Series B ($115M, Dec 2022, Spark Capital led), with a reported ~$34M follow-on in April 2026.

emerging

The question that decides it: Does Afresh's per-SKU perishable-ordering ML wedge survive once (a) Instacart's Carrot Enterprise AI bundles forecasting into the warehouse-pick, ads and smart-cart stack grocers already buy from Instacart, (b) RELEX, Blue Yonder and SymphonyAI retrofit LLM-era fresh forecasting into contracts the grocer already pays for, (c) Kroger-style vertical automation (Ocado-powered CFCs) handles ordering natively inside the fulfillment stack, and (d) the collapse of the Kroger-Albertsons merger triggers portfolio reshuffling that puts Afresh's single largest reference account into play? Falsifiable test for winning: company discloses ARR crossing $75M by end-2027, retains 2,200+ Albertsons store footprint at renewal, lands at least two additional top-20 US grocers, and ships genuinely usable center-store modules at 100+ stores. Falsifiable test for losing: an Instacart or RELEX bundled fresh forecasting feature displaces Afresh at one of the top-10 US grocers, or Afresh raises a down-round or sells for less than its $115M Series B total.

My take ★★ Very interesting ★ Interesting − Maybe ✗ Not interesting

Snapshot

Afresh sells a cloud ordering engine for perishable categories — produce, meat, seafood, deli, bakery, dairy — that replaces handheld-scanner and spreadsheet workflows with daily per-SKU, per-store recommendations trained on each retailer’s POS, inventory, waste and promotional history. Fresh is ~40% of US supermarket sales and the bulk of shrink, so single-digit waste improvements are enormous dollars. Afresh has raised ~$148M — $12M Spark-led Series A (March 2020), $115M Spark-led Series B (December 2022; Gunderson Dettmer; VC News Daily) at a reported ~$400-450M post-money — plus a reported ~$34M April 2026 follow-on (raising.fi). Its flagship customer, Albertsons, rolled the produce engine to 2,200+ Shaw’s, Jewel-Osco, Star Market and other-banner stores in 2024 (Produce Market Guide; AndNowUKnow, 2024).

Founding story

Schwartz, Fenner and Kuleshov met at Stanford. Schwartz did Management Science & Engineering, then a brief rotation at McKinsey before leaving to found Afresh in 2017. Fenner became CTO; Kuleshov — a Stanford CS Ph.D. who later took an assistant-professor seat at Cornell Tech within the Bowers College of Computing and Information Science (Cornell Bowers, 2024) — became Chief Scientist. In interviews Schwartz has described the unlock as realising that category managers were handed category-level forecasts when the real decision — how many cases of romaine hearts to order into store 1174 for Tuesday morning — is a per-SKU, per-store, per-day problem no one had systematized (Code Story, 2022). The founders ran regional-chain pilots in 2018-19, converted those into a $12M Series A led by Spark Capital in March 2020, and Schwartz was named to The Packer 25 class of 2022 (The Packer, 2022) — uncommon trade recognition for a Silicon Valley founder.

How it works

At a traditional supermarket, perishable ordering is done by a department manager who walks the aisle with a handheld scanner at dawn, eyeballs stock, flips through ordering guides, and places an order that lands via DC-to-store cross-dock 24-72 hours later. The forecast lives in their head; the system of record is the next day’s shrink. Afresh replaces that loop with a tablet app: the engine reads the retailer’s POS, inventory, waste logs and promotion calendar, and surfaces a recommended per-SKU per-store per-day order the manager can accept, nudge or override. Models are retailer-specific — trained on each grocer’s own history, accounting for weather, local demand, feature-and-display lift, holiday curves, DC lead times, and item-level yield loss. Execution data flows back; recommendations improve with tenure. The product also layers category-level inventory visibility and store-ops features (shrink root-cause flags, out-of-stock alerts, forecast-vs-actual dashboards) that give central merchandising a store-by-store view the handheld-scanner workflow never produced.

Product and business overview

Afresh for Produce is the original wedge and deepest module. Afresh for Meat, Seafood, Deli, Bakery and Dairy rolled out between 2022 and 2023, covering the full fresh perimeter (Food Market; Deli Market News, 2023). Afresh for Center Store, launched in 2024, extends the engine into shelf-stable packaged goods — the much larger category by SKU count but with easier forecasting and far more incumbent competition from Blue Yonder, SymphonyAI and RELEX (Grocery Dive; Progressive Grocer, 2024). Afresh for Foodservice targets non-retail buyers of fresh (foodservice distributors, QSR commissaries). All of these ride a shared engine; the department-specific product is a differently tuned model and workflow, not a separate codebase.

Business model and pricing

No published prices. Deals are enterprise annual SaaS contracts scoped by store count, department coverage and optional modules. Deployments are high-touch — six-to-twelve-month implementations with change-management work that typically costs the grocer more than the SaaS fee. Based on retail-SaaS comps, deals plausibly sit in the low-four-figures-per-store-per-month range per department, stacking across multi-department rollouts. The sales motion is reference-heavy: pilots graduate into full-banner rollouts only after demonstrable waste reduction (company-reported produce case studies cluster in a 20-40% range — not independently audited). Revenue is not disclosed; Growjo (2025) sits in mid-single-digit to low-double-digit millions, with upside as the Albertsons ramp completes.

Traction over time

A credible but concentrated climb: product that works, landmark reference customer, trailing commercial engine not yet obviously spreading the win across the top-10 US grocers.

Market analysis

US supermarket shrink ran ~1.6% of sales in 2023 on an industry base above $800B, implying a direct annual shrink pool above $12B; estimates of US food-retail food waste sit above 40 billion pounds a year (Worldmetrics; Gitnux; ZipDo, 2026, drawing on USDA-anchored sources). Fresh carries the bulk of that waste on a dollar basis and the thinnest raw gross margins before shrink — produce typically prints 32-35% gross that collapses toward the low 20s once realised waste is counted. Afresh is attacking a shrink-and-COGS leak large enough that modest percentage reductions fund the SaaS fee many times over. The handheld-scanner-and-spreadsheet status quo is finally being replaced, and the replacement will be either a specialist (Afresh), an incumbent planning platform (RELEX, Blue Yonder, SymphonyAI), an in-house build on Snowflake or Databricks, or a vertically bundled offering from Instacart.

Competitive intel

RELEX Solutions is the most direct threat — PE-backed, >500 retail customers, a dedicated fresh module, and 2024-25 announcements of accelerated subscription growth (RELEX, Jan 2025; Jan 2026). It already runs fresh at Rossmann, M&S Food and Franprix, bundling fresh into a supply-chain suite the retailer buys once. Blue Yonder (Panasonic) is the entrenched US incumbent; Albertsons itself has historically run Blue Yonder/JDA for broader planning, so the political case for a bundled Blue Yonder refresh is non-trivial. SymphonyAI sells CINDE Demand Performance Suite into the same category-manager buyer (Forrester Landscape 2025), attacking from the enterprise-planning layer. Shelf Engine, Afresh’s closest VC-era peer, took on forecasting risk itself (buy-back guarantee rather than SaaS), signed Kroger and Target pilots, and wound down in 2023 — the clearest cautionary tale in the category. Instacart Carrot Enterprise AI is the quiet strategic risk: existing fulfillment relationships with most of Afresh’s customer base, forecasting bundled with pick-pack-deliver and shopper-demand data Afresh does not have (Instacart, 2024-25). o9 Solutions, now public, sells demand planning up-market; relevant mostly to center-store deals.

History and evolution

Milestones are in the traction table above. Two inflections worth noting beyond that: the Kroger-Albertsons merger block in December 2024 (Axios; Just-Food, 2024) unwound the context Afresh’s largest customer was planning in, and headcount reset from the ~250 implied by 2023 hiring plans into a 150-250 band through 2024-25 — a quiet post-ZIRP recalibration rather than a disclosed layoff.

What people say

The case for. Trade coverage has been consistently positive on the product’s mechanical substance: Progressive Grocer, Grocery Dive, Produce Market Guide, Supermarket News and The Shelby Report have run partnership pieces emphasising sustainability wins and in-stock improvements across Albertsons, Heinen’s and others. Albertsons leaders — SVP of Fresh Dain Charette and former COO Susan Morris — have publicly endorsed Afresh (AndNowUKnow; The Produce News, 2024). Walter Robb’s involvement is substantive signal from someone who ran one of the most operationally serious fresh-grocery brands in the US.

The complaints. Public ARR is undisclosed; Growjo’s estimate (2025) is a fraction of what a $115M Series B at ~$440M needs to grow into. Albertsons concentration is both the most impressive asset and the most dangerous dependency, especially after the Kroger merger collapse (Axios; Just-Food, 2024). Shelf Engine’s 2023 wind-down is the cautionary reference competitors raise in every deal. Team Blind threads reference compensation and promotion-velocity concerns consistent with a company that priced itself for a 2021 environment and now has to grow into it (Team Blind, 2024-25). The ambient skepticism — that a specialist perishables SaaS cannot survive bundled competition from Blue Yonder, RELEX, SymphonyAI and Instacart indefinitely — is more felt than written.

Outlook: the open question

Afresh wins if, by end-2027, it has disclosed ARR crossing roughly $75M, held its Albertsons footprint through renewal, added at least two more top-20 US grocer logos, and shipped center-store at 100+ stores. It loses if an Instacart Carrot or RELEX bundled fresh forecasting feature displaces Afresh at a top account, or it is forced into a down-round or a sale for less than its $115M Series B total. The deeper question is whether vertical AI depth in one grocery sub-system is a durable company or a feature. Shelf Engine answered that one way. RELEX, Blue Yonder, SymphonyAI and Instacart are working to answer it the same way. Afresh’s bet is that fresh is operationally and model-architecturally different enough from ambient that specialist depth compounds, that Albertsons’ reference credibility pulls three or four more tier-1 US grocers across the line, and that center-store expansion gives it a credible single-platform story before bundled competitors arrive at the same answer from the other direction.

How to attack it

Four wedges are credible. Vertically integrated grocery OS: a startup (or Instacart, already positioned here) offering a single operating system for the grocer — ordering, assortment, pricing, pick-pack, ads, loyalty — with forecasting as connective tissue rather than a standalone product. The grocer buys one platform, not five; Afresh’s single-category SaaS is strategically naked against this. LLM-first category manager: collapses the category-manager workflow (promotion planning, assortment, forecast review, supplier negotiation) into a semi-autonomous agent rather than a tool pushing forecasts into the department manager’s tablet. Afresh pulls budget from the shrink line; a category-manager agent attacks the much larger merchandising headcount line. Dark-store OS: with Kroger-Ocado CFCs, Walmart Spark in-store pick, and the erosion of the store manager’s physical ordering workflow, forecasting lives natively inside the fulfillment layer; a startup building the OS for 2030-era dark stores folds Afresh’s wedge inside its stack from day one. Retailer self-build on Databricks/Snowflake: the stack to build competent per-SKU fresh forecasting in-house is cheaper every year; tier-1 retailers already own the underlying POS, inventory and waste data on their own data clouds, and a Databricks or Snowflake solutions partner could stand up a competent replica for materially less than Afresh’s multi-year contract. Documented weaknesses: no public ARR; Albertsons concentration magnified by the collapsed Kroger merger; a single-category story only recently extended to center store, where RELEX, Blue Yonder, SymphonyAI and o9 are already dug in; a 2022 valuation the 2024-25 cost reset implies has not been grown into; and the Shelf Engine precedent, which costs Afresh something in every grocer’s own risk committee.

Adjacent-segment play

The engine is category-general: any buyer with perishable, demand-volatile, lead-time-sensitive inventory is a potential customer. Foodservice distributors (US Foods, Sysco, Performance Food Group) are the most natural adjacency: per-SKU, per-depot, per-customer ordering at scale, and Afresh’s 2023 foodservice module was the earliest sign the company saw it. Convenience chains (Circle K, 7-Eleven, Casey’s) sell a growing share of fresh prepared and perishable food on much shorter cycles and with thinner store-ops bench strength than supermarkets; specialist perishable ordering would land well. Big-box perishables (Walmart, Target, Costco) is the biggest-dollar prize, but Walmart’s in-house build and Instacart’s Target relationship make it the most defended; Costco is a plausible niche. QSR commissaries and chain-restaurant central kitchens (Starbucks, Chipotle, Panera supply chains) forecast perishable raw-material need daily and would benefit from Afresh’s layer, though integration into kitchen-production planning is non-trivial. Comps already attempting adjacent variants: Toast Supply, BlueCart (B2B foodservice ordering), Crisp (CPG supply-chain data), and Snowflake-partner analytics shops working inside Sysco and US Foods. The adjacency Afresh is best positioned to lead on today is foodservice distribution — models, unit economics and customer-development muscle translate cleanly.

Sources and further reading

• Spark Capital Leads $115M Series B Financing of Afresh — Gunderson Dettmer, December 2022
• Afresh’s $115M Series B financing — Global Legal Chronicle, December 2022
• Afresh Technologies venture capital funding — VC News Daily, 2022
• Albertsons Cos. installs Afresh tech for produce in 2,200-plus stores — Produce Market Guide, 2024
• Albertsons and Afresh Announce Nationwide Rollout — AndNowUKnow, 2024
• Afresh expands AI technology into center store — Grocery Dive, 2024
• Afresh Expands into Meat, Seafood, Deli, and Foodservice — Food Market, 2023
• Heinen’s Partners with Afresh — Perishable News, 2023
• 2022 Packer 25 — Matt Schwartz — The Packer, 2022
• Volodymyr Kuleshov faculty page — Cornell Bowers College, 2024
• Compare Afresh vs RELEX — CB Insights, 2024
• RELEX Closes 2024 with Accelerated Growth — RELEX / BusinessWire, January 2025
• SymphonyAI AI Retail Planning Platform in Forrester Landscape 2025 — SymphonyAI, 2025
• Instacart AI Solutions for all grocers — Instacart, 2024-25
• Kroger-Albertsons merger blocked by Washington state judge — Axios, December 2024
• Afresh Technologies Raises $34 Million to Boost AI Platform (reported) — raising.fi (unconfirmed in primary sources), April 2026