Why are businesses slow to adopt AI?

The main reasons are price, data-retention and compliance concerns, complexity relative to need, and the integration effort of changing how work happens.

Are open models good enough for business use?

Yes, for most office tasks. Open models are now only months behind frontier models and match them on general work — drafting, summarizing, and data handling — at a fraction of the cost.

How much do frontier AI models cost businesses?

Anthropic's flagship Fable 5 runs roughly $10 per million tokens — about double OpenAI's GPT-5.6 Sol — and the data shows businesses reserving it for specialist work, not everyday tasks.

What's the fastest-growing category in business AI adoption?

Model-serving platforms offering open-source models — used by 6.1% of AI-adopting businesses in July 2026 and climbing month over month.

Why Businesses Are Slow to Adopt AI (and Open Models Are Winning)

Field Notes is our ongoing look at what's actually happening in AI and what it means for small businesses in Spring, TX and across Greater Houston. No hype, just the signal.

AI adoption is real — and it's slowing

More U.S. businesses are buying AI than ever. The Ramp AI Index (August 2026) tracks real business spend, and the headline numbers look healthy: 43.5% of U.S. businesses now pay for Anthropic subscriptions or tokens, and OpenAI sits at 39.7%.

Look closer, though, and the pulse is weaker. Month-over-month growth at both companies has slowed to a crawl — OpenAI added just 0.23 percentage points in July. New businesses are still starting with the big American labs, but existing adopters — especially the advanced spenders — are starting to look elsewhere.

The fastest-growing category in the data isn't a frontier lab at all: it's model-serving platforms that provide access to open-source models, now used by 6.1% of AI-adopting businesses and climbing month after month.

Why are companies slow to adopt AI?

The reasons show up clearly in both the spend data and the conversations happening in working teams:

1. Price. The gap between "good" and "best" is enormous. Anthropic's flagship Fable 5 costs roughly $10 per million tokens — about double OpenAI's GPT-5.6 Sol — and one month after launch it captured only 6% of business tokens, but 11.4% of Anthropic's dollars. Businesses vote with their wallets: performance beyond what the job needs isn't worth the premium.

2. Data handling. For any business with legal, compliance, or security concerns — which is most of them — data retention is a deal-breaker. If a model's prompts are kept for review windows measured in weeks, sensitive client data stays off the platform, period.

3. Complexity and overkill. Most office tasks — summarizing conversations, drafting decks, routing emails — don't need frontier ability. Teams describe flagship models as "a Ferrari when most businesses just need a van": more capability than the workflow will ever use, plus guardrails that make professional work feel unreliable.

4. Integration effort. Adopting AI means changing how work happens, not just buying a subscription. Slow adoption is often slow integration — the time cost of figuring out where AI fits beats the promise on the box.

Open models: the quiet alternative

The spend data points to a clear trend: open-source and open-parameter models are now only months behind the frontier — close enough for most if not all general office work, at a small fraction of the price.

That's why model-serving platforms are the fastest-growing slice of business AI. Companies are finding they can run drafting, summarization, data extraction, and even customer-facing automation on open models at costs that make the per-token math disappear from their budgets — and run them on infrastructure they control, which answers the data-retention objection entirely.

For a small business, this changes the calculus completely. You no longer need frontier pricing for office-grade AI. You need the right model matched to the right task — which is exactly the gap custom automation fills.

What this means for small businesses in Houston

Three practical takeaways from August's data:

You're not behind for being cautious. The slow pace of AI adoption is rational, not lazy. The businesses winning aren't the ones with the most AI spend; they're the ones with the clearest ROI per dollar.
Match the tool to the task. A $10-per-million-token flagship can't beat a $0.50 model at summarizing your inbox. Your money belongs where the automation is — renewals, proposals, data syncing between the apps you already use.
Control your data. Open models and self-hosted options give you the same everyday capability without shipping client information through third-party retention windows. For clinics, insurance agencies, and anyone holding sensitive records, that's not a feature — it's the deal.

The frontier race gets the headlines. The businesses getting value are the ones who noticed the race doesn't matter for their Tuesday.