FROM THE EDITORS
August 14, 2026
Yesterday we argued that Gemini is being built for a person with a phone, not a person with a business. The lifestyle concierge angle — booking dinner reservations, finding concert tickets, streaming music — is impressive, but it doesn’t pay an invoice, reconcile a P&L, or keep a client project on track.
Twenty-four hours later, Google dropped Gemini 3.7 Flash. And for the first time in a while, the announcement reads less like a consumer gadget pitch and more like a tool a business could actually justify paying for.
Here’s what changed — and what still hasn’t.
The price cut is real, and it matters
Google is offering 3.7 Flash at an introductory price of 0.75 per million input tokens and 3.75 per million output tokens — half the cost of 3.6 Flash. That’s not a marketing rounding error. For any business already using AI for content drafting, customer support triage, or light automation, this is a meaningful reduction in operating cost.
If you’re running a small agency, a freelance practice, or a lean in-house team, token pricing is your unit economics. Halving that cost while claiming better performance is the kind of move that shifts AI from an experiment to a line item you can defend in a budget review.
It actually got better at business tasks
The benchmarks Google published are unusually specific about workflows that matter outside a phone screen:
- AutomationBench jumped from 17.0% to 30.4%. That’s a near-doubling in the model’s ability to complete real-world business workflows.
- GDP.pdf — a test for processing complex documents — went from 22.0% to 34.0%. For anyone pulling data out of contracts, vendor quotes, or compliance PDFs, that’s a genuine improvement in accuracy.
- DeepSWE v1.1 coding performance rose from 49.0% to 65.3%. If your team uses AI for prototyping, internal tools, or web development, the first-pass code quality just got noticeably sharper.
These aren’t vanity metrics about beating a chess engine. They’re about the kind of work that fills a Tuesday afternoon in a real office.
The developer experience improved where it counts
Google’s own description of 3.7 Flash focuses on something we’ve been asking for: disciplined execution. The model is supposed to adapt better when it hits a roadblock, clarify intent when instructions are ambiguous, and put more effort into multi-step planning and tool calls.
In practice, that means fewer 3 AM debugging sessions because the agent looped itself into nonsense. Less manual oversight. Fewer retries. For a small business where “the person who knows the API” is also “the person who handles client billing,” that friction reduction is worth more than any benchmark.
Spark in Workspace: still a subscription play, but a sharper one
Gemini Spark — the 24/7 agent for Google AI Pro and Ultra subscribers — is now running on 3.7 Flash. The pitch is improved tool use across Gmail, Docs, Slides, and the rest of Workspace.
This still locks useful business functionality behind a consumer subscription tier. But if you’re already in the Google ecosystem, the upgrade means Spark can now do things like consolidate files across projects, draft status updates, and manage email workflows with fewer hallucinations. It’s not a replacement for a human operations person, but it’s starting to look like a credible intern that never sleeps.
What’s still missing
We’re not rewriting our position entirely. A few hard questions remain:
- The pricing is “introductory” through the end of the year.
That’s a four-month window. If you build a workflow around this cost structure and the price doubles in January, your unit economics collapse. Google has a habit of calling things “introductory” and then making them permanent, but the uncertainty alone should make any finance-minded founder cautious.
- The business features are still API-first, not app-first.
To get the real value — the agents, the document processing, the coding assistance — you need to be comfortable with Google AI Studio, the Gemini API, or the Enterprise Agent Platform. That’s a developer or integrator audience, not a small business owner who just wants the thing to work inside Gmail without reading documentation.
- The consumer DNA is still dominant.
The same blog post that announced 3.7 Flash leads with a 3D game demo, a parallax landing page, and a robotics training loop. The business PDF-to-dashboard example is buried halfway down. Google still thinks the sexiest way to sell AI is through spectacle, not through a quiet Tuesday afternoon where someone reconciles thirty invoices in twenty minutes.
The verdict
Gemini 3.7 Flash is the first Gemini release in 2026 that we’d recommend a small business actually test. The price is right, the document and workflow improvements are measurable, and the reduced need for hand-holding makes it practical for teams without a dedicated AI engineer.
But test it as a tool, not a strategy. Build one workflow. Measure the output. Watch the pricing page like a hawk. And don’t let the 3D game demos distract you from the actual question: does this save you time, money, or sanity on the work you do every day?
Yesterday, we said Gemini was built for a person with a phone. Today, Google made a credible case that it might also be built for a person with a spreadsheet. We’ll see which version of the story sticks.
Want to talk about where AI actually fits in your business?
Get in touch — we build workflows that work.
Leave a Reply