N Catalyst.ai
Book a demo
Home / Blog
Blog
Forty problems worth solving.
Specific, hard situations from the industries we work in: defence, healthcare, legal, pharma, manufacturing, recruitment and contact centres. Each one is a real problem, not a capability tour.
No posts in that industry yet.
When the aircraft is newer than its paperwork
Modification state lives in service cards, amendment logs and a register that stopped being reconciled three years ago. Here is how to make the tail number the source of truth.
ReadThe part that only one supplier makes
Single-source exposure is usually discovered when the supplier stops answering. Finding it in the contract record first is an exercise in reading paperwork nobody has opened in years.
ReadReading a technology transfer dossier nobody indexed
A transfer package arrives as thousands of pages in someone else's structure, in someone else's language of part numbering. Making it usable is the first engineering task, before any engineering starts.
ReadThirty days before the due date
Service dates are written in prose, on cards, in a format designed for a technician rather than a scheduler. Turning them into something that can raise a reminder is the whole problem.
ReadAnswering a 400-page tender without missing a clause
Tender compliance fails on omission, not on argument. The hard part is establishing the full set of things you are required to respond to.
ReadMore sensor log than anyone will ever watch
Uncrewed sorties generate review backlog faster than analysts can clear it. The useful question is not how to watch it all, but how to find what warrants a human.
ReadThe discharge summary that arrives after the follow-up
Summarising an admission is not the hard part. Making the summary complete, attributable and timely enough to be clinically useful is.
ReadFinding the prior that changes the read
A current study is read against history. When the history is thirty documents across eight years and two systems, the relevant prior is often the one that is not found.
ReadWhen the protocol changed and the notes did not
Clinical pathways are updated centrally. Documentation habits are updated locally, slowly, and unevenly. The gap between the two is invisible until somebody audits it.
ReadThe appointment reminder that actually reaches someone
A text message is cheap and easy to ignore. A call works and costs. Automating the call changes that arithmetic, but only if it handles what the patient actually says back.
ReadA trial master file that can survive an inspection
TMF completeness is judged against an expected structure. The work is establishing what is present, what is missing, and what is filed in a place the inspector will not look.
ReadDisclosure at a scale no team can read
Proportionality is the legal test. Defensibility is the practical one. A review that cannot show how it reached its decisions is not finished, however fast it ran.
ReadWhat did we actually agree to, across 900 contracts
Every individual contract is readable. The portfolio is not. The obligations that bite are the ones nobody has read since signature.
ReadThe intake form that decides the next six months
Most of what a matter needs to be scoped correctly arrives in the first bundle. Very little of it is in a structured field.
ReadBreak clauses, rent reviews, and the dates nobody diarised
A property portfolio's risk is concentrated in a small number of dates, most of which are derived from clauses rather than recorded anywhere.
ReadWhen the regulation changes, which documents go stale
The hard part of a regulatory change is not understanding it. It is finding everything you have written that now says something different.
ReadBatch record review that does not gate release
Review by exception is the goal. Getting there means deciding what counts as an exception, from a record that was designed to be read in full.
ReadHas this deviation happened before
Investigating a deviation without the history risks treating a recurring systemic issue as a one-off. The history exists. It is just not searchable in the way the question needs.
ReadThe SOP that references a machine you retired
Procedure estates decay quietly. The failure is discovered during an audit, by someone reading a procedure that describes equipment that left the site years ago.
ReadQualifying a supplier on evidence, not on reputation
A qualification pack is a collection of assertions supported by certificates. Checking that the support actually exists is the work.
ReadThe spare that went out of production in 2019
Obsolete spares stop production lines. The decision between a redesign, a last-time buy and making the part is usually made without the information it needs.
ReadFour thousand applicants and nine interview slots
At high volume, the screen is decided by whoever reads the first hundred applications. Everyone else is filtered by proxies that correlate with background rather than ability.
ReadScreening engineers without burning engineer time
Technical screening is expensive because it consumes the time of the people you are hiring to relieve. That cost shapes the process more than anyone admits.
ReadCampus season in a two-week window
Campus hiring compresses a year of interviewing into a fortnight. Capacity, not selectivity, decides who gets seen.
ReadTwo interviewers, two different questions, one decision
When candidates are asked different questions, the hiring decision compares interviewer impressions rather than candidate evidence.
ReadThe twelve questions that make up most of your queue
Most inbound volume is a short list of questions asked repeatedly. Handling them well is a different problem from handling them at all.
ReadThe follow-up that never gets made
Outbound follow-ups are always the first thing dropped when inbound volume rises, which is exactly when they matter most.
ReadReviewing one call in a hundred, or all of them
Quality assurance by sampling tells you about the average. The calls that create risk are, by definition, not average.
ReadPrinting a part whose drawing is on microfilm
Making a legacy part is rarely blocked by the printing. It is blocked by establishing what the part actually is.
ReadSelling a design instead of a part
Designers reach customers who want an object, not a file. A marketplace that prints and fulfils changes what a design is worth.
ReadThe paperbook is due on Monday
A paperbook is not a PDF of the file. It is a paginated, indexed, internally consistent document, and assembling one by hand is the junior's weekend.
ReadObjected at the filing counter
A registry objection costs a day, sometimes a limitation period. Almost all of them are for defects that were visible in the file before anybody left for court.
ReadThe case that does not say what it is cited for
A proposition travels from a headnote into a note into a draft, and by the time it reaches the court nobody has read the paragraph it supposedly comes from.
ReadTwo criminal codes, one live matter
Offences committed before the new codes came into force are tried under the old ones. For years to come, practice means holding both in your head at once.
ReadFour thousand notices and one panel of advocates
Recovery at portfolio scale is the same document produced thousands of times with different facts in it, and every one of them still has to be right.
ReadDates and synopsis, from a record nobody has read end to end
The list of dates is the first thing the court reads and the last thing anyone has time to build. It is also the only part of the paperbook that is pure extraction.
ReadThe deposition that is a photocopy of a photocopy
The record arrives as faded certified copies, handwritten depositions and orders in three languages. Most document tools silently give up on exactly this material.
ReadThe gap in the chain of title
A title opinion is a chain of documents with no missing links. Finding the missing link is a reading exercise, and the documents are long, repetitive and in several languages.
ReadGetting ready for the DPDP Act without a privacy team
Most Indian businesses that process personal data have no privacy function at all. Readiness for them is not a programme; it is a set of documents that have to exist and be true.
ReadPrivilege at the scale of a disclosure exercise
In a large review, the cost is not the privileged documents. It is the hundreds of thousands that are not, and still have to be looked at.
Read