Every fuel lab has someone like this. They have run the octane engine for twenty years. They know it needs a longer warm-up in winter, that readings drift when the room humidity climbs, and which adjustment brings it back. They know the second position on the distillation unit runs slightly warm, that the flash point tester needs an extra cleaning after a biodiesel blend, and that the sulfur analyzer reads a little high for a few runs after maintenance. They can look at a copper strip and give the rating that matches the reference card, every time.
None of that is written down. It does not need to be, as long as they are there.
Why each instrument is its own case
Two units of the same model, running the same ASTM method, do not behave identically. Age, repair history, the room they sit in, and the products they usually see all leave a mark. Experienced analysts learn these habits the hard way, through retests, failed proficiency rounds, and late nights. Over time they build a private manual for every instrument: what normal looks like, what the early warning signs are, and what to do before a result goes wrong.
That private manual is valuable. It is also fragile, because it exists only in people's heads.
The risk most labs do not price in
When a senior analyst leaves, retires, or is simply on holiday during a busy week, the lab feels it quickly:
- More retests and slower turnaround. Newer staff cannot recognize a drifting instrument, so problems are caught after results are produced instead of before.
- Inconsistent results. Subjective tests, such as visual ratings, depend on who is reading them. Customers notice when the same product gets different answers.
- Audit exposure. Accreditation requires documented procedures and evidence that each analyst is competent on each method. If real practice differs from the written SOP, an assessor will find it.
- Long, expensive onboarding. A new chemist can learn the method in days. Learning the instruments takes months of shadowing, and only works if the right person has time to teach.
- Disputes that are hard to defend. When a cargo result is challenged, the lab needs to show why the result is right. "Our senior analyst always does it this way" is not evidence.
For a lab owner, this is a business continuity issue. A large part of your lab's reliability may be sitting with two or three people, without a backup plan.
Training is the control, and it needs better material
Most labs already train well on methods. The gap is instrument-specific knowledge. A strong training program covers both: the standard method, and how each instrument in your lab actually behaves while running it. It also keeps a clear record of who is qualified on which method and which instrument, and when they were last assessed.
The difficulty is producing that material. Senior analysts are the busiest people in the lab. Asking them to write manuals in their spare time rarely works.
How AI helps capture what your people know
Much of this knowledge is already partly recorded, scattered across logbooks, maintenance notes, deviation reports, retest reasons, proficiency test results, and emails. AI can work with what you already have and make it usable.
- Gather the scattered record. It reads past logbook entries, deviation reports, and maintenance notes for each instrument and groups recurring issues and fixes.
- Interview your experts efficiently. Short structured sessions with senior analysts are turned into clear written notes, which the analyst then reviews and corrects.
- Draft instrument guides. For each instrument, it drafts a practical guide: known behaviors, warning signs, checks before a run, and what to do when something looks off. Each point links back to the record it came from.
- Answer questions on the bench. A newer analyst can ask why a result looks unusual on a particular unit and get an answer drawn from your own records, with sources, rather than guessing or waiting for the one person who knows.
- Keep SOPs and training current. When practice and procedure drift apart, it drafts the proposed SOP update and the matching training material for QA to review and approve.
What stays with your people
The knowledge belongs to your analysts, and they remain the authority on it. AI drafts. Senior staff confirm what is right. QA approves what becomes procedure. Nothing changes in a controlled document without the people who already own that decision. The result is a lab where knowledge is written down, reviewed, and available to everyone, not held by a few.
How to start
Ask one question: if your most experienced analyst left next month, which instrument would worry you most? Start there. Collect what is already recorded about that instrument, spend a few hours with the person who knows it best, and produce one reviewed guide your newer staff can use tomorrow. Then do the next instrument, ideally while the people who know it are still in the building.