Answer-first guides on the problems we get hired to solve — healthcare AI and LLMs on regulated data, revenue cycle automation, denial management, payer rate analytics, and data platform migrations. Written to be useful first and findable second.
Four layers matter: a warehouse, ingestion, transformation in version control, and one dashboard people open. Eight things you can skip and the symptom that tells you when to stop skipping, when Postgres stops being enough, and the licence trap in the BI layer.
Published August 20, 2026
The tooling in most tutorials has been retired — DMA, the Azure Data Studio extension, and DMS classic are all gone. What replaced them, how to choose between SQL Database, Managed Instance, and a VM, the Managed Instance link's version cliff, and the objects no tool carries across.
Published August 18, 2026
Health plans publish every in-network negotiated rate under a 2020 federal rule, and almost nobody uses it. The seven-step method for going from a payer's index file to a defensible market percentile: which schema fields carry the analysis, the five cleaning rules that kill ghost rates, and what the files still cannot tell you.
Published August 15, 2026
Sort the work into rules, models, and language tasks and automate in that order — most of the savings need no AI at all. Which stages automate, what the CAQH data says it is worth, what silently breaks, and what actually happens to a billing team.
Published August 13, 2026
The four jobs AI actually does in denials — cluster to root cause, rank by recoverable dollars, predict before submission, draft the appeal — plus the three decisions it should never make, and what the KFF and Premier data say about how much of this is preventable.
Published August 11, 2026
The four arguments your own claims and remittance data can make — rate gap, systematic underpayment, volume, and dollar impact — how to pull the allowed amount out of an 835, and the contract language that quietly cancels a rate win.
Published August 6, 2026
The % of Medicare method, step by step: express each contracted rate as a share of the Medicare allowable for the same code and locality, volume-weight to rank your payers, and find the codes where a raise moves the most dollars.
Published July 23, 2026
Azure OpenAI is HIPAA-eligible under Microsoft's BAA — but eligible isn't compliant. What the BAA covers, what stays your responsibility, the settings that silently break it, and how it runs on real PHI in production.
Published July 8, 2026
Leaving a GoDaddy reseller tenant buys full admin, Entra ID, and an Azure runway. The reseller ceiling, the direct-tenant payoff, the cutover step by step, the data-loss failure mode — and how a mailbox truncated in the cutover was recovered in full from the 30-day soft-delete window.
Published July 8, 2026
Databricks' Lakebridge auto-converts most Teradata SQL — but silently skips what it doesn't recognize. What converts, what you rewrite by hand, the tool's failure modes, and the sqlglot workaround for when the CLI won't install.
Published July 6, 2026
The denial codes that drain the most revenue — CO-16, CO-197, CO-11, CO-50 and more — with a table of root causes, which are preventable, and where each one gets fixed in the revenue cycle.
Published July 6, 2026
Consumer ChatGPT has no BAA, so PHI is off-limits. What a compliant LLM deployment actually requires — a signed BAA, training disabled, access controls, audit logging — and which platforms qualify.
Published June 30, 2026
If you are weighing AI on regulated data, untangling denials, or planning a migration, the fastest path is a conversation. Bring the specifics; leave with next steps and an estimate the same business day.
Book a 30-minute intro call Prefer email? clayton@quantsolvent.co