Stop running the business from spreadsheets and inboxes

Who this is for

This playbook is for owner-operated companies where the Monday number, the invoice, and the CRM row can all be “the customer” and still not match. It is also for the operations lead who has been asked to “get the data in one place” and is being sold a platform before anyone named an entity.

If the pain is that people cannot find the current rule or the latest deck, start at operational knowledge. Knowledge and data sit next to each other, but a trusted article is not a trusted invoice. If the pain is a weekly pack rebuilt from four exports, read weekly reports after you lock the definitions — a brief on top of two cash numbers is still an argument.

Symptoms

If several of these are true in the same month, this is a live operational leak — not a tooling preference.

• Leadership asks “what is the number?” and two competent people send two spreadsheets.
• The CRM, the billing tool, and a personal workbook each claim to be the customer list.
• A report is late because someone is reconciling columns that should have been the same entity.
• Inbox threads are treated as the current status because the system of record is stale.
• A vendor demo starts with a warehouse or a chatbot before anyone named which entity would move.
• Someone proposed an AI layer that would answer from “all of our sheets” without saying which sheet wins.
• A prior “migration” copied last year’s mess into a new tool and both copies are still live.

What done looks like

Done is a working operating change, not a purchased seat or a dashboard nobody opens.

• Each argued-about entity has one named source of truth and one named owner.
• A written definition exists: what the entity is, what is excluded, and which system may write it.
• One entity has completed a pilot: new work writes to the chosen source, and the dual copy is retired.
• Reports and billing for that entity read the chosen source only.
• No AI, chatbot, or dashboard is allowed to answer from two live copies of the same entity.
• A failed or dirty migration is written down as a non-repeat: no second big-bang until the pilot holds.

Source of truth per entity, not “all the data”

“Consolidate the data” is how companies buy a platform and still argue on Monday. Narrow it. An entity is a thing the business already has a fight about: who is a customer, what is an open invoice, which jobs are active, what cash landed this week. Pick those fights. Leave the rest of the company’s files alone.

A source of truth is a write rule, not a feeling. One system is allowed to create and change the record. Other systems may read it or show a copy, but they do not get to invent a second customer. If two systems can both create “the same” customer, you do not have a source of truth. You have a race.

Owners matter as much as systems. A CRM with no person who will reject a duplicate is a folder. A spreadsheet with a named owner and a weekly check can be a temporary source of truth. Do not confuse “we paid for a system” with “someone keeps it true.”

Run one entity as a pilot

Choose the entity that already hurts a decision this month. Cash and invoices are common. Active jobs are common. The whole customer universe is a poor first pilot because it is large and political. You want a boundary you can finish.

Write the definition in plain language. Example: “An open invoice is a record in the billing tool that has been sent and is not marked paid or void. Spreadsheet tabs and inbox PDFs are not open invoices.” That sentence is the pilot. Everyone who touches the weekly pack has to use it.

Then change the write path. New invoices are created only in the chosen system. The dual spreadsheet stops receiving new rows. Historical rows can sit as an archive with a date and a “do not update” note. The pilot is done when a week passes and nobody needed the second copy to answer a live question.

Billing depends on this. If delivery, CRM, and billing disagree on the engagement, you will invoice a story nobody can defend. Read catch scope creep before it never hits the invoice once one engagement entity has a single picture. Do not auto-invoice across a dual customer list.

Kill the dual source of truth

The dual spreadsheet is the real product. People trust it because they can see the cells. The official system is “wrong” because nobody kept it. Consolidation that leaves both live is a rename. You have to retire one write path.

Retirement is a behavior change, not a delete on day one. Hide the second file, move it to an archive folder, revoke edit access, and put a note at the top: “Closed. Open invoices live in billing.” If someone still needs a number from history, they can read. They cannot extend the argument.

Inbox copies die the same way. If status lives in a thread, the system of record is already losing. Put the status in the chosen system and answer the thread with a link, not a new shadow table. This is tedious. It is also the work.

Do not put AI on two spreadsheets

A model that can see both copies will sound fluent and still pick a side at random. That is not insight. That is a confident coin flip. If you want answers, the source of truth has to exist first. If you want a knowledge layer for policies and how-tos, that is an operational knowledge base— trusted documents with owners, not a mash of ledgers.

The same rule applies to weekly reporting. Stop rebuilding weekly reports by hand assumes you can lock a definition. If two exports still disagree on cash, automate the brief after the entity is one sentence, not before. A push brief that cites two sources without a winner trains leaders to ignore the pack.

Ranking belongs here. “Buy a warehouse” and “buy a chatbot” are purchases. Rank AI work that pays before either one. Low data readiness is a cleanup, not a model.

A worked pilot, without a warehouse

Imagine a service firm where Monday’s cash number comes from a bookkeeper’s workbook, the billing tool, and a deposit column the owner edits in email. Do not assign a savings figure. Name the entity: cash landed this week. Write the definition: “Cash this week is deposits marked cleared in the billing tool. The workbook and the email column are copies.” Name the owner: the bookkeeper. For two weeks, the Monday brief cites only the billing tool. The workbook is moved to an archive folder with a closed date. If a deposit is missing, write it in the billing tool, not in a new cell in the sheet.

That is a finished pilot. It does not migrate customers, jobs, or last year’s invoices. It does not invite a warehouse vendor. It produces one sentence leaders can reuse: we know where cash is allowed to live. The next entity — open invoices, or active jobs — earns a turn only after this sentence holds when the bookkeeper is out.

If the firm also argues about who the customer is, do not fold that fight into the cash pilot. Two entities at once is how big-bang thinking sneaks back in. Finish cash. Write the customer fight on a later card.

The failure is a dirty big-bang migration

The usual disaster is familiar. A vendor promises one platform. Someone exports everything. Duplicates, stale customers, personal notes, and three versions of “paid” land in the new system. The team keeps the old spreadsheet “just in case.” Six weeks later both copies are live and the new tool is blamed.

Do not migrate dirt because the contract started. Clean the pilot entity first, or migrate a thin slice: active customers only, open invoices only, jobs from this quarter only. History can wait in a read-only archive. Completeness is not the same as truth.

Big-bang also fails politically. Every department has a sheet they believe. Moving all of them in one weekend creates a coalition against the new source. A finished pilot creates a proof: this entity, this week, one answer. That is how the next entity earns a turn.

What people usually get wrong

They start from the tool. Warehouses, CDPs, and “AI data layers” are later builds. If you cannot name the entity and the owner, you are buying a place to put the argument.

They keep the dual spreadsheet because it is “faster.” It is faster until the Monday fight. Speed that produces two answers is not operational speed.

They treat a one-time import as consolidation. An import without a write rule is a snapshot. Snapshots rot. The source of truth is the path for next week’s new row.

They ask AI to reconcile the mess. Reconciliation is a definition and an owner. A model can help draft the definition. It cannot be the owner.

DIY vs hire

DIY one entity. This month, pick the fight that already shows up in a leadership meeting. Write the definition. Name the source and the owner. Stop adding rows to the second copy. Run one report or one invoice list from the chosen source. If the team cannot live without the dual file, the definition is not done — rewrite the definition before you buy a platform.

Hire help when the argued entities span sales, delivery, and finance and nobody owns the cross-cut, or when a migration is already on the calendar without a pilot. The $997 assessment is a live consultation and a written analysis. You leave with two to four improvements you can run yourself — an entity list, a definition, a retire plan for one sheet — plus a later roadmap quoted only if you want us on joins, integrations, or a warehouse. Implementation is not in the fee.

Controls before automation

Recommendations that touch customers, money, contracts, private data, or systems of record need human review, limited permissions, a log, and a fallback. Do not automate a broken path because the tool is ready.

• No AI, chatbot, or dashboard answers from two live copies of the same entity.
• No write-back to a system of record until the field map is written and a person can undo it.
• Do not start a company-wide migration while a dual spreadsheet is still the Monday source.
• New tools inherit the access the human already had — not a wider key to “see everything.”
• Keep a fallback: if the chosen system is down, a dated export with an owner can answer — a second living sheet cannot.
• If two systems disagree, stop the automated report or invoice. Fix the definition first.

How this sits next to the other playbooks

This is the answerable room in the fourteen mid-market AI asks: make one entity true before you ask software to speak. Knowledge bases cover documents and rules. Weekly reports cover the brief. Billing covers work that should become an invoice. Ranking decides whether a dirty data picture is the blocker in front of all three.

If you want that order written down with you, book the assessment. Bring the two spreadsheets that already disagree. Leave with one entity, one owner, and a retire date for the second copy.