Bittner AI

How it works

Every step either produces an answer or stops and tells you why.

Most of what goes wrong in an intake automation is not a bug. It is a place the system stayed quiet when it should have spoken. This is the chain, and what each step does when it fails.

The chain

Seven steps, and what each one does when it fails.

  1. The submission is written down before anything else runs

    The raw payload is logged the moment it arrives, ahead of any external call. If verification is down, if the scoring step fails, if the practice-management API is having an afternoon — the enquiry is still recoverable by hand. Nothing is lost because a later step broke.

    Fails: never silently. The record exists first.
  2. A bot gate stops junk before it costs anything

    A hidden field no person can fill in. Submissions carrying it stop here, before verification or scoring spend anything on them.

    Fails: toward letting it through to a human, not toward blocking a real client.
  3. Phone and email are verified

    The number is checked and normalised, and the line type recorded. The email is checked for deliverability. This is where most junk is caught — but a real person who mistypes their own address lands in the same bucket as junk, so unverified is never treated as fake.

    Fails: to a named person at your firm, who is the safety net for a typo.
  4. Your contact list is pre-screened

    Against the opposing party named in the enquiry, and separately against the lead's own email to catch a repeat inquirer. If a genuine flag comes back, every client-facing message is held and you are told who matched, by name.

    Fails: loudly. A search that cannot run stops the record — it never reports zero.
  5. The matter is scored against your practice

    In scope or out, practice area, county, seriousness, completeness, a rating and the reason behind it. The scoring is written for your practice areas and your county footprint — which is the step that knows a Berrien County matter from an Indiana one.

    Fails: to human review. Never to an automated rejection.
  6. It routes, five ways

    Verified and qualified gets a reply and a booking link. Verified and out of scope gets a courteous decline — wording you have read and approved. Unverified goes to your named person. A conflict flag sends the client nothing and alerts you. A scoring failure does the same.

    Silence toward the client is the correct output for an unresolved conflict.
  7. The contact is created and you get one email

    The record lands in Clio with the intake detail in custom fields — source, practice area, verification status, line type, summary, score, the reason, and the conflict flag. Your email answers one question: should I call this person back.

    Client mail sends from your account, with your disclaimer language on every send.
The part that matters

Deciding in advance what happens when a step fails.

Automation for a professional services firm is mostly the discipline of deciding, ahead of time, what happens when something breaks and who finds out. The code is the easy half.

Four rules come out of that, and they are not negotiable in an engagement:

  • No safety gate that fails silently toward the client. A blocked enquiry that receives nothing and is told nothing means a person who tried to hire a lawyer simply never heard back. A flag you see beats a gate that swallows.
  • “Couldn't check” never renders as “nothing found.” On a conflict search those are opposite answers. An error that stops the run and parks the record is correct. Quietly continuing with a zero is not.
  • Unverified is not fake. Typos, catch-all domains and unknowns all come back not-valid. Someone at your firm owns that inbox by name — not us, because a safety net that lives with the contractor disappears when the engagement does.
  • You read every client-facing sentence aloud before go-live. An automated decline goes out under your name and your bar number. Whether an automated decline is acceptable at all is a professional conduct call, and it is yours, not ours.

On the word “conflict”

What runs is a pre-screen against your contact list. It does not search matters, related parties, or opposing parties recorded in matter fields, and it never returns a cleared status. We hold that line in the system, in proposals, and on this page — because a vendor who lets “pre-screen” drift into “check” in a sales conversation is telling you what the rest of the engagement will be like.