The batch is finished. The paperwork decides whether it ships.
Between the last process step and release sits work your team carries by hand. Botable builds and runs the AI layer that takes that stretch back, working around your systems and never inside them. Release keeps the owner it has today.
Start with one gapAn open item is never just paperwork.
A deviation still open. A change waiting on its full picture. Training that has not reached the night shift. Behind some of them is a batch that cannot move. Behind all of them is a person keeping the work moving.
The batch is physical. The hold is not.
A batch can be made, sampled, and sitting in quarantine while a deviation, an investigation, or an open change control keeps it from moving. What frees it is record work: items reviewed, reconciled, and closed by your people.
The revision goes effective in a day. Training lands shift by shift.
A single SOP revision can create required training for operators, QC analysts, warehouse staff, and managers, each group on its own schedule, across shifts and sites. The requirement is set and the completion is gathered person by person, but chasing it has been someone's unwritten job.
The impact is judged in pieces. The decision needs it whole.
A proposed change can touch the process, the analytical method, the specification, and the filing. Each function can judge its own piece. Someone still has to pull the pieces into one picture and walk it to every desk, and the change holds until they do.
In GMP work, close enough is a deviation.
The procedure that governs a step is controlled, current, and often nowhere near the person performing it. Working from memory is not an option, and leaving the work to go check is an expensive one. The answer has to come to the work, carrying the revision and the passage that lets a person verify it on the spot.
The vendor question is now in writing.
The February 2026 Computer Software Assurance guidance covers software used in production and quality management, and it addresses AI and machine learning used in that role. It expands recommendations on vendor oversight and clarifies when 21 CFR Part 11 applies.
The AI layer stops where the record begins.
Official record
The systems you already own keep the record.
Batch records, deviations, procedures, training assignments, change controls: each stays in the system that governs it, in the state your team approved. The layer holds no records of its own and creates no second version of yours.
Work around it
The AI layer does the work.
It finds the current procedure and its citation, keeps training follow-through moving, and lays a change's full reach out in one place. Open items surface where reviewers already look. Everything it produces is an input to your review, not an entry in your record.
Human decision
Your team signs off. The layer never does.
Release or hold, disposition, approval, closure: each carries the name and the signature of someone on your team. The layer's work ends one step earlier, with everything assembled in front of that person.
Pick one gap. Keep every system.
Each one is narrow on purpose: a single piece of hand-carried work, built on the systems you keep exactly as they are.
When completion has to be gathered shift by shift
Training Notification Orchestration
It takes the requirement to each affected person on their own schedule, keeps the reminders on cadence, and escalates to a manager while there is still time to act. Where things stand stops being a question your team answers by hand.
When the governing procedure is rooms away from the work
eQMS Search & Retrieval
An operator or analyst asks in plain language and gets the governing passage back, cited to its document and revision, current as of today. An answer that cannot be verified is not treated as an answer.
When the impact of one change sits in five functions
Change Order Management
It gathers what a proposed change touches, across process, analytical method, specification, and filing, and delivers the whole of it before the review, not during it. Each function still makes its own call on its own piece.
Bring us the paperwork your batches wait on.
We map the work, build the AI layer alongside the systems already in place, and stay to run it as procedures change, shifts rotate, and the next batch comes off the line.
Start with one gap