AI Automation Can Change More Than the Task You Automate

What changed

On September 24, the Blue Cross Blue Shield Association (BCBSA) released an analysis of commercial inpatient hospital claims from its member plans, finding that more intensive billing added an estimated $942 million in costs over two years, compared with a 2023 baseline. Of that, $653 million came from claims where a secondary diagnosis, a separate condition found in addition to the main reason for the hospital stay, pushed the bill into a higher-paying category.

BCBSA ties the shift to the rise of AI coding and revenue-cycle tools. More than 60% of hospital systems now use AI systems, including ambient scribes that listen to patient visits and draft notes, that scan lab results, notes, and records to surface conditions a clinician may not have flagged for billing on their own. Across BCBSA's member plans, the share of hospital stays coded as "medically complex" rose from 37% at the start of 2023 to 40% by the end of 2025.

The finding that got attention: BCBSA compared the coding trend to treatment data and did not find a matching rise in the treatments typically given for those newly added conditions. In other words, more stays are being coded as more complex, but the care itself isn't showing the same increase.

The honest nuance

This is an insurer trade association's own analysis of its own claims data, and BCBSA has a direct financial interest in describing hospital coding as inflated: higher coding intensity means higher payouts to hospitals. That doesn't make the numbers wrong, but it's a reason to read the framing carefully rather than take "AI is driving up costs" at face value.

BCBSA itself was careful to say the pattern doesn't prove any individual claim was fraudulent or that AI alone caused the entire increase; it described the coding-versus-treatment gap as suggestive, not proof of wrongdoing. The American Hospital Association pushed back, arguing that patients really have gotten older and sicker over this period, and that hospitals may simply be documenting conditions that were always present but previously went unrecorded. Both explanations, more complete documentation and coding creep, can produce the same number: more complex codes without more treatment.

Why it matters for your business

The specific story is about hospital billing, but the underlying lesson travels well beyond healthcare: an AI tool can do its narrow job accurately, extracting a real detail from a document, flagging a real pattern, and still change a financial outcome as a side effect, simply because it's more thorough or more consistent than the manual process it replaced.

A lead-scoring tool that's better at spotting buying signals will route more leads to your top commission tier. A support assistant that's quicker to say "this qualifies for a refund" will approve more refunds. An invoicing tool that catches every billable line item you used to round down or skip will raise your average invoice. None of that is a bug in the AI. It's the AI working exactly as intended on a narrow task, while quietly moving a number nobody was watching.

What to actually do

Pick one AI-assisted workflow you already use that touches a financial outcome, pricing, discounts, refunds, commissions, lead scoring, approvals, and pull 15 to 20 of its recent decisions. For each one, work out what the pre-AI process would have produced for the same input, using your old rules, old templates, or your own judgment before you had the tool.

Then compare the rate of the costly or high-value outcome, not just whether the task got done faster. If an AI-assisted refund workflow is approving refunds at twice the old rate, or a lead-scoring tool is putting 30% more leads in your top tier than a human reviewer would have, that's the number to investigate, even if every individual decision looks defensible on its own.

From Kindloom Labs

That kind of audit starts with knowing where AI is actually touching your workflows in the first place. The free checklist Small Business AI Use Case Checklist organizes common AI use cases by business function, marketing, operations, and customer service, which makes it easier to spot the ones that affect money and deserve a closer look.

Sources

• BCBSA Analysis: How AI Coding Tools Affect Healthcare Costs (Blue Cross Blue Shield Association)
• Hospitals' use of AI coding tools cost BCBSA plans $942M more for similar care: analysis (Fierce Healthcare)