writing
building in public: what we ship, what breaks, what we learn.
ai receptionist for law firms: what it should answer, what it hands off
A comparison of answering services, generative voice bots, and deterministic intake systems for law firms, and where each one crosses into unauthorized legal advice.
missed call text back for a law firm: what to send, never promise
a missed call text back for a law firm is an intake receipt, not legal advice. here's what the message should say, and the three ways it gets firms in trouble.
prior authorization automation: chase the paperwork
prior authorization automation that chases the paperwork, not the clinical call: a model reads, code checks, a human signs off before anything is sent.
is ChatGPT HIPAA compliant? what a practice can put in it
standard ChatGPT is not HIPAA compliant. what a practice can safely run through it, where PHI has to stop, and how to build the gap.
ai automation companies: platforms, agencies, or consultants
platforms, agencies, and custom builds solve different problems. here's how to tell which one your regulated workflow actually needs.
ai automation for a business that can't afford a wrong answer
ai automation for small business works when the model reads and a deterministic gate decides. the architecture that keeps mistakes from shipping silently.
hipaa compliant automation for the back office, not the exam room
most healthcare ai targets the exam room. hipaa compliant automation is safer and more useful in back-office intake, where fixed rules govern every step.
mass tort intake: grading volume without grading merit
a working method for mass tort intake where a model extracts facts and deterministic code qualifies leads, so nothing is deciding merit but a lawyer.
human on the loop vs. human in the loop for regulated work
human on the loop watches after the fact. human in the loop approves before anything ships. for regulated work, only one of those survives a false positive.
what a workflow automation consultant actually does in the first two weeks
a workflow automation consultant should spend the first two weeks building nothing. here's what those weeks contain, and what you hold at the end.
referral management software, for a practice whose referrals arrive by email
before a practice buys referral management software, it should look at where referrals arrive. for many it's a shared inbox, and that changes the answer.
personal injury lead generation companies vs. leads your firm owns
personal injury lead generation companies sell leads; the alternative is a pipeline your firm owns. what each buys you, and the step both get wrong.
legal intake services vs. an intake system your firm owns
legal intake services can mean rented people, rented software or a system you own. the question that sorts them: who writes the rules, and who keeps the record.
what law firm automation actually covers, and shouldn't
law firm automation works when a model reads and stages the work, and a deterministic rule or a human makes every final call. here is where the line sits.
medical billing automation without the handover of decisions
medical billing automation can speed up claims without letting a model decide them. the fix is splitting extraction from decision, with a trail either way.
how to choose an ai automation consultant you can't evaluate
you cannot judge ai output directly. here is how to evaluate the process, the infrastructure, and the incentives of an ai automation consultant instead.
AI readiness assessment: a test for what not to automate
an AI readiness assessment methodology that tells you which decisions the model should never make, not just what to buy. a test you can run yourself.
medical practice automation: where the audit trail starts
accuracy is the wrong first question for medical practice automation. here's why the audit trail has to come first, and what it actually looks like.
legal intake automation that grades facts and never merit
legal intake automation should extract and structure facts for review, never score a case's merit. here's the mechanism and why the line matters.
questions to ask ai vendors before you sign anything
the questions to ask ai vendors when one pitches your firm: auditability, fallback, data residency and exit terms.
how to run AI on client data without a compliance headache
a practical way to put AI on client data without handing it decision power: split reading from deciding, own the infrastructure, keep a human in the loop.
why an audit trail matters more than accuracy
a 99% accurate model that can't show its work is a bigger liability than a slower system you can audit line by line. here's the actual split.
how to avoid a hostage situation with your automation vendor
a working system, held in someone else's cloud, is a hostage negotiation waiting to happen. here is how to structure ownership before you sign.
human in the loop automation: AI extracts, it never decides
human in the loop automation works when the AI extracts facts and a person or a rule makes the call. reviewing a model's decision is theater.
an AI that grades facts vs one that grades merit
AI can extract facts from a file reliably. Assessing legal merit is a different task entirely, and confusing the two is how firms get burned.
what an ai audit trail actually looks like
most vendors log model inputs and outputs and call it an ai audit trail. that tells you what happened, not why. here is the difference.
deterministic vs probabilistic ai: code makes the call
deterministic vs probabilistic ai in regulated work: the model should read and route, and only deterministic code should decide.
ai for regulated industries: the work, not the decision
ai for regulated industries works when the model touches the work but never the decision: a method that keeps every decision in auditable code.
how to explain an automated decision to a regulator
a regulator does not need to trust the model. they need a decision they can trace to a rule. here is how to build that trail.
what business processes can be automated (and what can't)
what business processes can be automated? the test isn't repetitive vs. rare, it's fact vs. judgment. six steps to find the real boundary.
parallel ai agents won't save you (the part the demos skip)
throwing 300 parallel ai agents at a problem isn't the unlock. the unlock is what brackets the swarm: before and after.