Consulting Slide Library Best Practices: Taxonomy, Ownership, and Reuse

A slide library works when finding an approved slide is faster than rebuilding it. That is the entire success criterion. Everything below is in service of that one sentence.

Most libraries fail it. I have used several that were technically comprehensive and practically useless, and the pattern was always the same: I searched, found four versions of the thing I wanted with no way to tell which was current, gave up, and built it again. Multiply that by every consultant in the firm and you can calculate what the library actually cost you.

So let me be specific about what makes the difference.

The three failure modes

Nobody owns it. Content goes in and nothing takes it out. Two years later, half the library is wrong and everyone knows it, so nobody trusts any of it. Staleness is contagious in a way people underestimate: one out of date slide teaches a user that the library is unreliable, and they will not search again for months.

The taxonomy was built by filers, not searchers. Someone organised by project, by year, by author, by department. Perfectly logical for putting things away. Useless for the actual query, which is always some version of "I need a slide that explains our approach to X for a client in Y."

Retrieval loses to reconstruction. If searching takes six minutes and rebuilding takes eight, people will search. If searching takes six minutes and copying from last week's deck takes ninety seconds, the library is decorative. Speed is not a nice to have here. It is the mechanism.

Structure by question, not by origin

Organise around what a user is trying to answer.

Three levels deep at most. Category, subcategory, slide. A seven level taxonomy feels rigorous during design and produces empty branches within a year, because nobody wants to make five decisions to file one slide.

The metadata that carries the weight

Folder position tells you almost nothing. The fields do the work.

The last field is what turns a review conversation into a lookup. When a partner asks whether the market figure on page nine is current, the answer is a record with a date, not an archaeology project.

Ownership that actually holds

The word "ownership" gets used loosely, so here is what it has to mean in practice.

An owner is a named individual who is accountable for the accuracy and currency of a specific set of slides. They receive review reminders. They can retire content. Their name appears on the record. And critically, the time this takes is acknowledged rather than assumed to be free.

That last point is where most libraries quietly die. Knowledge maintenance gets assigned to whoever has a quiet week, and nobody in professional services has a quiet week. If the review work is not somebody's actual job, with actual time, it does not happen, and no amount of process design compensates.

A workable model: practice leads own methodology and capability content and delegate specific sets to named senior people. Marketing owns positioning. Legal owns mandatory pages. Engagement owners own their own credentials. Knowledge management owns the taxonomy, the standard visuals and the health of the whole thing.

Review cycles by content type

Not everything ages at the same rate. Match the cadence to the decay.

• Market and benchmark data: quarterly. Numbers with no date on them are worse than no numbers.
• Capability and positioning: every six months, or immediately when the offering changes.
• Methodology: annually, and whenever the method itself is updated.
• Credentials: annually, and whenever a client relationship or permission changes.
• Mandatory and legal pages: whenever the wording changes, driven by legal rather than by a calendar.
• Standard visuals: annually, mostly for brand and template compliance.

Then the rule that makes review cycles real: past the review date, a slide becomes unavailable rather than merely old. Not hidden in a warning banner people click through. Unavailable.

This feels harsh and it is the single highest impact policy in the whole article. A slightly stale slide will always get used, because using it is less work than updating it. A slide the system will not serve gets updated, or it gets retired, and either outcome is better than the current one.

Curate ruthlessly

Library size is inversely related to library value past a surprisingly low threshold.

A curated set of one hundred to three hundred genuinely current slides beats a repository of eight thousand, every time. Trust is the asset. Every stale slide a user encounters costs you future searches, and those losses compound.

So: retire aggressively. If a slide has not been used in eighteen months, it is telling you something. If four slides say roughly the same thing, pick one and kill three. If nobody will own a slide, that is a decision, and the decision is to remove it.

The resistance you will meet is that somebody might need it someday. They might. They will also cope, and the cost of that occasional inconvenience is much lower than the cost of a library nobody believes.

Design retrieval for the real query

People do not search the way librarians expect. They search in the language of the client problem, under time pressure, with partial recall.

Four things that measurably help:

Search content, not just titles. The slide someone needs may be titled "Operating model transformation" when they are searching for "target operating model" or "TOM".

Support synonyms in the controlled vocabulary. Map the variants to one concept rather than letting four spellings fragment your results.

Show currency in the results. Review date and approval status visible before the click. This one change removes most of the "is this current" hesitation that sends people back to copying old decks.

Respect permissions at retrieval. Client cleared credentials must not surface to someone building an unrelated pitch. This is enforcement, not review, and it needs to work through every surface including search, export and any AI interface.

What changes when AI is retrieving

Two shifts, and both push in the same direction.

Metadata matters more, folders matter less. An AI system can search across a well tagged library instantly regardless of structure. What it cannot do is distinguish an approved current slide from a stale draft unless a field says so. Approval status, review date and confidentiality level become load bearing rather than nice to have.

Curation matters more, not less. People assumed AI would make large messy libraries workable. The opposite happened. A system retrieving from a library with four versions of the same slide will confidently return one of them, and it has no way to know it picked the wrong one. Volume without governance produces confident wrongness at scale.

The rule that follows: an AI system retrieves approved records or reports a gap. It never composes a substitute from fragments of the library. That single behaviour is what separates a useful assistant from a liability.

Measuring library health

Six numbers tell you almost everything:

• Search success rate: the share of searches ending in a slide being used.
• Time from search to insertion, compared with time to rebuild.
• Share of produced decks drawing on library content.
• Percentage of slides past their review date.
• Slides with no owner, which should be zero.
• Retirement rate, because a library that never removes anything is not being curated.

Watch the first number over time. Search success is the leading indicator for everything else. When it falls, people stop searching, and once they stop it takes a long time to win them back.

Where offgen fits

In offgen, the slide library is part of the Company Brain rather than a separate repository. Slides carry owner, approval status, review date, market, language and confidentiality level. Retrieval runs inside existing permissions. Brand governance keeps reused slides consistent with the current master, so an approved slide from eighteen months ago does not arrive with an old logo. Identifiers survive into the generated deck, so any slide in a draft can be traced to its record.

More on how this fits practice workflows is on our consulting page.

The question I would ask about your current library: when a consultant needs a methodology slide at nine in the evening, do they search it or rebuild it? Everyone in your firm already knows the answer. It is worth asking them.