This moment has many doors in: you bought a company and the seller handed over the server. You succeeded a parent in the family business. You are the new controller after a sudden departure, the successor trustee, the person who now runs the buildings. However you arrived, the situation is the same: an archive built by someone else, holding obligations you have not met yet, organized by a logic you do not know, and too large to read. The instinct is to clean it up. The instinct is wrong, and this guide is the order of operations that works instead.
The short version
Secure and back up everything exactly as it stands, because the structure itself is information. Find the documents with dates attached before anything else, since obligations do not pause for your transition. Build a one-page map of what lives where instead of trying to absorb the contents. Write the twenty questions you actually need answered and hunt those answers, citing where each was found. Purge nothing for ninety days, and after that only against a written retention decision. If the person who kept the archive is reachable, one hour walking the folders with them is worth weeks of solo reading.
First, freeze it
Before anything is opened, moved, or judged, the archive gets copied and backed up as it is. Not because everything in it matters, but because you cannot yet tell what does, and because the arrangement is the previous keeper's map of the operation. Folder names, what sits next to what, which drawer held the leases: that structure is often the only finding aid in existence. Reorganize before you understand it and you haven't cleaned the archive, you've burned the index.
Freezing has a technical half and a human half. The technical half is a copy that lives somewhere the predecessor's credentials cannot reach and the office cannot accidentally overwrite, made before anyone opens a drawer or a share. The human half is telling the people who touch these records that nothing gets thrown out, reorganized, or tidied for now, because the most common way inherited archives lose material is a well-meaning assistant clearing space in week two. Say it once, in writing, to everyone with access.
Hunt the clocks, not the contents
The dangerous documents are the ones with dates. Leases and the notice windows for their renewals. Insurance policies and their renewal dates, because a lapse discovered at claim time is the expensive version. Loan agreements, their covenants, their maturities. Contracts that renew themselves unless someone objects by a deadline that may be next month. Permits and licenses with expirations. Warranties still running. The tax and filing cadence the business has been keeping without you. The first pass through an inherited archive is triage by clock, and everything date-bearing goes onto one calendar with the document it came from noted beside it.
Sort what you find into three buckets as you go. Things that expire, where missing the date ends something: policies, permits, licenses, registrations. Things that renew themselves, where missing the date starts something you did not choose: service contracts, subscriptions, equipment leases with evergreen clauses. Things that come due, where a payment or a filing has a deadline attached: notes, tax filings, regulatory reports. The three behave differently under pressure, and the second bucket is the one that quietly costs money, because nobody gets a reminder for an obligation that continues by default.
Give the calendar a lead time, not just a date. A renewal window that closes in March is a February task at the latest, and often a January one if a decision needs a quote or a board vote first. Enter the notice deadline as the event and the underlying expiration as a note beneath it, because the deadline you can actually miss is the notice.
Build the map, not the library
You do not need to know what every document says. You need to know what kind of thing lives where. A one-page inventory does it: each top folder or drawer, what it appears to hold, roughly what years, who created it. This takes an afternoon, and it converts the archive from a wall into a searchable territory. It is also the page you will hand your accountant, your lawyer, and your eventual buyer, all of whom will ask the same question: what do you have?
Keep the map descriptive and refuse the urge to grade it. "Four boxes, vendor files, roughly 2009 to 2016, kept by the office manager" is useful. "Four boxes of junk that need sorting" is a judgment made before you know anything, and it tends to become self-fulfilling. Where you genuinely cannot tell what something is, write down what it looks like and move on; the unknowns are a short list you will resolve in the course of the twenty questions, not a project of their own.
Write the twenty questions
An archive earns its keep by answering questions, not by being tidy. Write the twenty you actually have. What is the insurance deductible. When does the anchor lease renew and how much notice. Where is the operating agreement and what does it say about exactly this transition. Which vendors have contracts and which have habits. What did we pay for the roof and who warranted it. Then hunt the answers, and record where each was found: document, page, folder. That answer sheet, with citations, becomes the institutional memory the archive was hoarding, and building it teaches you the territory faster than any cleanup would.
Write the questions before you go looking, because questions invented while searching tend to be the ones the archive can already answer easily. The useful list is the one that reflects real exposure: money that leaves on a schedule, obligations with counterparties, anything a lender or an insurer would ask about, and the handful of things you would need within an hour if something went wrong tonight. Twenty is a working number, not a rule; ten sharp questions beat forty vague ones.
Two answers deserve their own line on the sheet: the ones you could not find, and the ones you found twice with different content. The first list tells you what to ask the previous keeper. The second tells you which documents have superseding versions floating around, which is the failure that produces confident wrong answers years later.
Purge nothing for ninety days
Every inherited archive contains boxes that look like junk, and some of them are. But record retention obligations attach by record type and jurisdiction, tax years stay open, disputes reach back, and the box of apparent scrap paper is occasionally the one with the signed amendment in it. The rule that protects you: nothing leaves in the first ninety days, and afterward things leave only against a written retention decision, not against an impression. Storage is cheap. The document you destroyed is the one the lawyer will ask for.
When the ninety days are up, the decision is written and general, not made box by box. Retention decisions belong to record types, with a stated period and the reason behind it, and they get made with your accountant and your attorney once so they can be applied a hundred times. Anything with an active dispute, an open tax year, or an ongoing relationship attached comes out of the purge regardless of age. Record the decision itself in the archive; the next person to inherit this will want to know why a category ends where it does.
Interview the previous keeper
If the person who built the archive is available, or their longtime assistant is, an hour walking the top folders together is the highest-leverage hour of the whole transition. You are not asking them to explain the business; you are asking where things live and why. Record it. The map in someone's head has a way of being exactly the part that never made it onto paper.
Ask questions that surface habits, since habits are what the filing system encodes. Where did things go when they did not fit anywhere. What did you keep on paper on purpose. Which vendor or client has an arrangement that is not written down. What was moved, purged, or reorganized in the last few years, and why. Who else has copies. The answers rarely change what you keep; they change how fast you can find it, which is the whole point of the exercise.
Digitize the working set, and keep the paper that matters
Everything your twenty questions touched is the working set, and it earns scanning first, into the same folder structure the paper keeps. Scanning makes the archive searchable and answerable; it does not retire the paper. Keep originals of the documents whose value is the instrument itself: notes, titles, certificates, anything sealed or notarized, anything a counterparty could demand in original form. One labeled box of originals, everything else answered from the scan, is the stable end state.
Scan quality matters more than scan speed here. Text-searchable output instead of flat images, the original folder names preserved as you go, and file names that say what a document is and roughly when, since a decade of files called Scan_0042 recreates the problem you were solving. Multi-page documents stay whole; a lease split across nine separate files is technically digitized and practically useless.
Work in the order the twenty questions set: the documents that answered them first, then the categories most likely to be asked about next, then everything else as time allows. An archive that is thirty percent scanned along the lines of actual demand is more useful than one scanned front to back to the year 2004.
The traps
Reorganizing first, which destroys the finding aid. Trusting labels, because the folder marked Insurance was last accurate in some other decade; spot-check before relying. The miscellaneous boxes, which absorb exactly the documents that fit no category and therefore include the unusual ones that matter. Assuming a digital copy exists somewhere, when the scanner by the desk was for outbound mail. And in estates and family businesses, the quiet mixing of personal and business paper, which is a sorting problem and sometimes a legal one, worth flagging early to the professionals handling the transition.
Two more worth naming. The confident summary, where a spreadsheet or a memo describes what the documents say and everyone stops opening the documents; summaries drift from their sources and inherit nobody's corrections. And the single point of access, where one person holds the credentials, the keys, and the knowledge of where things are, which is precisely the arrangement that created your situation in the first place.
The corporate spine
Inside every business archive is a small set of documents that are not about operations at all; they are the entity itself, and the professionals arriving to help will ask for them first. The operating agreement or bylaws, and every amendment. The minute book, however thin. Formation documents, the federal tax identification letter, state registrations and qualifications, ownership records and any buy-sell agreements among owners, and the licenses and permits the business operates under. In a transition, these documents govern the transition: what authority transfers, who signs, what the operating agreement says about exactly the event that just happened. Locate the corporate spine in the first week, copy it, and put it where the lawyer and the accountant can reach it, because until they have it, much of what they do for you is guesswork with a disclaimer attached. If pieces are missing, and in long-running family businesses they often are, the gap itself is a finding worth surfacing early while reconstruction is still possible.
Reconstruction is usually possible and usually annoying. State agencies can reissue formation documents and good-standing certificates, the federal tax identification letter can be replaced, and prior counsel or the accountant often has the only complete minute book in existence. Start those requests in the first weeks, because they run on other people's timelines and they tend to become urgent exactly when a lender, a buyer, or an insurer asks for the file.
The digital estate
The modern archive has a second half that fits in no drawer: the domains, the email accounts and who administers them, the online banking tokens, the payroll and accounting logins, the customer lists inside a CRM, the subscriptions quietly auto-renewing on a card whose statement you have not seen yet. Inventory it like the paper: an accounts map listing each system, what it holds, who has access, and how access transfers. Two entries deserve urgency. Email, because password resets for everything else flow through it, and because a predecessor's mailbox is frequently the only archive of the agreements nobody printed. And the domain and website registrations, because an expiration missed during a transition takes the business offline in the most public way possible. Where credentials died with the predecessor, most major platforms have account recovery or succession processes; they are slow, they demand exactly the corporate spine documents above, and starting them early is the difference between an inconvenience and an outage.
Take ownership before you take inventory of features. Domains, email, and the primary financial logins should sit under an account you control, billed to a card you control, with recovery routed to an address that survives the transition. Where a vendor account is in the predecessor's personal name, that is a transfer request to make now, not a discovery to make during an outage. Write the accounts map down in one place and treat it as a document in the archive, since it is the finding aid for everything that has no drawer.
The people records are different
Somewhere in the archive are employee files, payroll records, benefits enrollments, and possibly medical information, and these are not documents to spread across a conference table while you get oriented. Employment records carry their own retention obligations and their own access restrictions, which vary by jurisdiction and record type, and mishandling them creates liability independent of anything else in the transition. The practical rule: identify where the people records live, confirm they are secure, restrict who touches them to the people who must, and route questions about them through counsel or the payroll provider, and let curiosity wait. The same caution extends to anything in the archive about customers that carries privacy obligations. This is the one category where the twenty-questions method yields to a narrower discipline: know where it is, keep it locked, ask before reading.
The practical version is separation. People records live in their own locked location, digital or physical, with access limited to the people who administer payroll and benefits, and they do not travel into the general scanning queue with everything else. If the archive mixes personnel files into operational folders, and small businesses often do, pulling them out is worth doing early and worth documenting as done.
The first ninety days, as a shape
The pieces above arrange themselves into a rhythm worth writing down. The first thirty days are for securing: the backup, the freeze on purging, the corporate spine located, the digital estate inventoried, the clock documents on one calendar. The second thirty are for understanding: the one-page map, the twenty questions hunted to cited answers, the hour with the previous keeper. The third thirty are for deciding: what the retention obligations actually are, written down; what the working set is, scanned; which originals go in the labeled box; what, if anything, leaves, against a written decision; impressions don't qualify. Ninety days is also roughly when the archive stops being an inheritance and starts being yours, which is the quiet goal of the whole exercise: not a clean archive, but an owner who can answer questions from it, with the source cited every time.
The rhythm assumes interruptions, because there will be some. If a deadline surfaces in week two, it jumps the queue and the rest slides; the sequence is a priority order, not a schedule. What should not slide is the first thirty days, since securing the archive and finding the clocks are the two steps whose cost rises fastest when deferred.
When the archive is also evidence
One situation overrides everything above, including the ninety-day rule, by making it stricter: an actual or reasonably anticipated dispute. Contested estates, partner disagreements, claims arriving with the transition, litigation the predecessor was managing quietly. Once a dispute is on the horizon, preservation duties can attach to exactly the records a cleanup would touch, and destroying them, even innocently, even on a schedule, can carry consequences far beyond the value of the shelf space. The rule collapses to one sentence: if any dispute is pending or foreseeable, nothing is destroyed and nothing is altered until counsel has said so in writing, and the scope of what counsel says to hold is read broadly. This is also a reason the freeze-and-back-up step comes first in every version of this process, because a complete copy made on day one is protection in both directions: proof of what existed, and freedom to work with the copies while the originals rest untouched.
The same instinct applies to routine deletion that runs without anyone deciding anything. Email retention rules, backup rotations, and document management policies that purge on a schedule will keep running through a dispute unless somebody stops them, and "the system deleted it automatically" is a poor answer to give later. When counsel puts a hold in place, ask specifically which automated processes need to be suspended, and confirm in writing that they were.
The insurance file, first among equals
Among the clock documents, insurance deserves to be pulled first, because it is the one whose failure mode is silent until the worst day. Locate every policy in force: property, liability, auto, workers compensation, umbrella, and whatever specialty coverage the business carries. Record carrier, policy number, period, and broker for each, and get the renewal dates onto the calendar before anything else from the archive gets read. Then make the call the transition makes necessary, because many policies carry notice requirements around changes in ownership or control, and coverage questions are settled far more comfortably before a claim than during one; the broker of record is the right first conversation and will usually reconstruct anything the file is missing. While you are in the file, pull the claims history too. It is part of the risk picture you just inherited, it affects renewal pricing, and an open claim being quietly managed by the predecessor is exactly the kind of thing archives hide.
Read what the policies actually cover while you are in there, since inherited coverage reflects a business that may no longer exist. Property limits set against values from years ago, liability limits set before the current work mix, a professional or cyber policy nobody renewed, or coverage for an activity the business stopped doing and is still paying for. None of this is a document problem exactly, but it is visible only from the documents, and the transition is the moment anyone actually looks.
The same drawer usually holds the adjacent instruments worth logging in the same pass: surety bonds, letters of credit securing leases or utilities, and any personal guarantees the predecessor signed, each with its expiration or release conditions noted, because these are commitments of the business that outlive the person who made them and rarely announce themselves.
Decode the filing system before judging it
Every archive has an organizing logic, and inheriting one means learning it before you grade it. The common schemes are few: by year, by project or job, by counterparty, by document type, or the archaeology of several eras layered as keepers changed. Sample three or four folders from different shelves and infer which scheme governs where, because the scheme tells you where things will be: in a by-counterparty system the roof warranty lives under the roofer's name, in a by-year system it lives in the year of the roof. Write the inferred scheme onto the one-page map, including where the eras change. And respect the miscellaneous drawer while refusing to trust it, since misc is where every scheme sends the documents that fit no category, which makes it simultaneously the least organized and most surprising cubic foot in the room, and a reasonable place to spend one focused hour early.
Whatever scheme you find, adopt it for anything you add. New documents filed the predecessor's way keep one system with one logic; new documents filed your way create a seam at the date you arrived, and seams are where things get lost. If the scheme genuinely does not work, change it deliberately later, in one pass, with the old structure preserved as a copy. What you want to avoid is the accidental hybrid nobody designed.
Frequently asked questions
I just bought a business. What should I do with the old records?
Back everything up as it stands, purge nothing for ninety days, and triage by clock: leases, insurance, loans, auto-renewing contracts, permits. Then build the one-page map and the twenty-questions answer sheet.
Should I reorganize the files?
Not yet, and probably never. The existing structure is the previous keeper's map and often the only finding aid there is. Keep it, make it searchable, and let the answer sheet do what the cleanup was supposed to.
What should I look for first?
Date-bearing documents: renewal windows, policy expirations, loan maturities, filing deadlines. Obligations do not wait for you to finish reading.
How long do records need to be kept?
It varies by record type and jurisdiction, and many categories carry multi-year obligations. Until you have mapped which apply, keep everything; afterward, purge only against a written retention decision.
Do I keep the paper originals?
Scan everything for daily use; keep originals of instruments: notes, titles, certificates, sealed documents, anything a counterparty could require in original form.