Count the funded AI agent startups from the last 18 months. Decagon, Sierra, Maven AGI - customer support. 11x - sales outbound. Aisera - enterprise IT and HR service. Harvey - legal. Glean - knowledge search.
Every one of them is customer-facing. Or lawyer-facing. Or knowledge-worker-facing. Someone, somewhere, is talking to a user and an AI is in the middle of that conversation.
Meanwhile, the thing that actually consumes the most hours inside most businesses sits almost untouched by AI startups: the admin pile. The supplier emails. The approvals. The forms. The compliance records. The weekly reports. The vendor certifications. The reconciliations. The thousand little things the owner or ops lead does between the software.
Why?
Why VCs ignored back office
Three reasons, roughly:
- Back office is unsexy. Sales demos of "autonomous CX agent resolving 80% of tickets" look amazing in keynote decks. Sales demos of "correctly approved 42 supplier invoices overnight, no discrepancies" don't have the same pop.
- Back office is per-business. Customer support is roughly similar across businesses - answer questions, resolve issues, escalate. Back-office work is idiosyncratic: every firm has its own supplier list, approval thresholds, forms, compliance requirements. Harder to sell a generic product into.
- Back office is under-measured. The owner of a 30-person business doesn't have a CFO-approved budget line for "admin in the middle." It's invisible cost. Hard to pitch ROI against an invisible cost.
All three are true. And all three are changing.
Why the market is opening now
LLMs crossed a threshold. Reading an email, matching it to a PO in a different system, applying a threshold, routing approvals, filing the signed copy - all of that now works well enough to automate with real reliability. 18 months ago, it didn't.
Owners got tired. Post-pandemic, labour costs are high and good admin hires are hard. Owners have been carrying more middle-work than they've carried in 20 years. The pain is real and the alternative hires are expensive.
The integration layer matured. Your supplier emails land in Gmail or Outlook. Your POs live in Xero or QuickBooks. Your compliance docs live in Drive or SharePoint. All have APIs. All are addressable. The plumbing exists.
Why the opportunity is huge
There are ~30 million small-to-medium businesses in the US and Australia alone. The median one has 5-50 disconnected tools and an owner doing 10-20 hours of middle-work per week. Replace that with AI and you're looking at a category that could eat multiple hundreds of billions in software and labour spend over a decade.
Compare that to customer support - a real market, but a narrow one. A fraction of businesses have enough ticket volume to justify a CX AI spend. Almost every business has admin.
What the winner looks like
The winning product shape is not another AI app you log into. It's an operations hub - a layer above your stack that reads from everything, writes back to everything, and runs the admin between your tools. Owner describes the business once. AI scaffolds the workflows. Platform runs them forever.
This is what Sprigr is built for. The quiet, unsexy, biggest opportunity in AI.