11 min read
It is 9:14 on a Monday morning and your VP of Sales is already in a spreadsheet.
Not the CRM. A spreadsheet. Because the CRM does not know how your company actually sells. It knows how Salesforce thinks companies sell. Those are different things.
Your VP pulled the pipeline data into a warehouse last night. Your RevOps analyst — the one you are paying $165,000 a year — spent Friday afternoon building a Python forecast because the native one does not match your sales motion. Your territory logic lives in a Google Sheet that three people understand and one person maintains. Your quoting rules live in a different Google Sheet. Your AI call-intelligence platform writes summaries that nobody bothers pasting back into the system of record because the system of record is not where work happens anymore.
It is where work gets reported. After the fact. By people who resent doing it.
And every January, someone signs the renewal. $1.2 million. Maybe $1.8 million if you are mid-market with aggressive per-seat pricing. Plus the two administrators. Plus the implementation consultant you brought back for the third time. Plus the integration middleware that breaks every time the vendor ships a release you did not ask for.
You know this. You have known this for a while.
How You Got Here
I am not going to pretend the CRM was always a mistake. It was not.
In 2012, you needed a canonical place for accounts, contacts, opportunities, pipeline stages, activity history, and permissions. You needed something your CFO could audit, your board could reference, and your sales managers could use to pretend they were inspecting pipeline when really they were just sorting by close date and hoping.
Buying a mature SaaS platform was the rational move. You did not have the engineering talent to build it. You did not have the tooling. You did not have the time. The platforms were good enough. They gave you structure when your sales process had none.
That was the trade. You gave up control in exchange for structure.
The problem is that your company grew up. Your sales motion got specific. Your pricing got complex. Your territories got weird. Your renewal economics diverged from your new-business economics. Your board started asking questions the CRM was never designed to answer.
And instead of admitting the platform no longer fit, you built around it. You added the warehouse. You added the BI layer. You added the Python model. You added the spreadsheets. You added the middleware. You added the RevOps team whose entire job is translating between the system you pay for and the business you actually run.
That is not a technology stack. That is a confession.
The Part Nobody Wants to Say Out Loud
Here is what your organization has already proven and nobody wants to write on a slide.
If your revenue team exports data out of the CRM to do the real work somewhere else — and yours does — you already admitted three things. The CRM is not sufficient. The important logic lives outside it. Your business needs a custom operating layer to function.
You built the escape hatch years ago. You just kept paying rent on the building you escaped from.
So now ask the adult question.
If the business already depends on custom logic outside the CRM, why is the CRM still organizing your operating model? Why are your sellers clicking through pipeline stages that a product manager in San Francisco designed for everyone and no one? Why are your managers reviewing forecasts in a format that does not match how your company closes? Why is your board package three hops downstream from a vendor’s object model instead of one step from your own data?
Why are you still paying for this?
“But the Data Model Is Complex”
No it is not.
This is the part that makes CRM vendors nervous when I say it in a room full of executives who actually think about it for thirty seconds.
Your customer data model is accounts, contacts, opportunities, pipeline stages, activities, contracts, and revenue events. That is it. It is important data. It is not a difficult data model. Your RevOps team could sketch it on a whiteboard in twenty minutes. Your data engineer could implement it in Postgres in an afternoon.
The CRM vendor spent fifteen years convincing you that this model is proprietary magic that only their platform can manage. It is not. It is a relational database with a permissions layer and a UI on top. You already have the relational database — it is the warehouse you export into every night. You already have the permissions layer — it is whatever IAM system your company runs. The only thing you are actually renting is the UI.
And your sellers hate the UI.
That is what $1.2 million a year buys you. A front end your team works around. An admin console that requires a specialist. An integration surface that fights you every quarter. A release cycle you do not control. A pricing model that scales with headcount regardless of value. And a contract renewal process designed to make switching feel more painful than staying — which is not the same thing as being good.
Four Engineers and Twelve Weeks
Three years ago, what I am about to say would have sounded irresponsible.
Say it anyway.
A small team — four strong product engineers using agentic development tooling — can now build a company-specific revenue operating system in twelve weeks. Not a prototype. Not a slide deck. Not a six-month architecture phase followed by a twelve-month buildout followed by a migration nobody wants to manage.
Working software. Deployed. Used by real sellers. Generating real pipeline data. On your infrastructure. With your data model. Matching your sales motion.
Week one. Your account model, your opportunity schema, your pipeline stages — modeled to match how your company actually sells, not how a vendor’s product manager thinks companies sell. A clean data layer in your warehouse. Basic CRUD. Auth. Permissions.
Week four. Your sellers have a working UI. It is not pretty yet. It does not need to be pretty yet. It matches their flow. The pipeline view shows what they care about. The opportunity record captures what matters to your business. Activity logging happens automatically from email and calendar. Nobody is manually entering call notes into a text field.
Week eight. Your forecasting model — the one your RevOps analyst built in Python, the one that actually works — is native. It is not downstream. It is not in a separate BI layer. It runs on the live data. Your managers see the forecast that matches reality, not the forecast the vendor’s algorithm hallucinates from incomplete CRM entries.
Week twelve. AI-generated account summaries from every call, email, and meeting. Deal risk identification before the manager asks. QBR prep that writes itself from the evidence trail. Missing stakeholder detection. Pricing anomaly flags. Forecast integrity checks that compare what the rep says with what the data shows. Tasks that open automatically when pipeline hygiene drifts.
That should keep CRM vendors up at night. A custom system with embedded AI does not just replace what the CRM does. It does things the CRM structurally cannot do. The vendor cannot build intelligence around your specific sales motion because they do not know your sales motion. They know a generic abstraction of everyone’s sales motion. You know yours.
Let Me Do the Math for You
Your CRM costs you more than you admit. Let me add it up the way your CFO should be adding it up but probably is not.
Per-seat licenses. Call it $1.2 million a year for a mid-market company with 200 sellers. Maybe more. Salesforce Enterprise is $165 per user per month before add-ons. Add CPQ, add the analytics cloud, add the AI features they are now charging extra for, and you are north of $200 per seat per month. For 200 sellers, that is $480,000 a year just in base licensing. But nobody runs base. The real number with add-ons, premium support, and the features you actually need is closer to $1.2 million.
Administrators. You have at least two. More likely three. They cost you $130,000 to $160,000 fully loaded. Call it $420,000.
The implementation consultant you keep bringing back because the last customization broke something. $150,000 a year if you are lucky. $300,000 if you are honest. Split the difference at $225,000.
Integration middleware — the thing that moves data between the CRM and your warehouse, your marketing automation, your billing system, your support platform. License cost plus maintenance plus the engineer who babysits it. $200,000.
RevOps headcount dedicated to translating between the CRM and reality. At least one full-time person. $165,000 fully loaded.
And then the part nobody puts on a spreadsheet. The workflow drag. The reporting latency. The forecast that is always three days stale. The pipeline review that takes 90 minutes because the data has to be manually reconciled before the meeting. The board package that requires a two-day sprint every quarter because the CRM’s reporting does not match how your board wants to see the business.
Add the hard costs. $1.2M + $420K + $225K + $200K + $165K. That is $2.2 million a year. Conservatively.
Now price the alternative. Four senior engineers at $225,000 fully loaded. That is $900,000 a year. Cloud infrastructure for a bespoke internal tool serving 200 users — $60,000 a year on the high end. AI API costs for the intelligence layer — $40,000 a year at current pricing, and it is getting cheaper every quarter.
$1 million a year. Total.
You save $1.2 million in year one. You save $2.2 million every year after that because the team that built the system maintains it — and they are building new capabilities instead of managing vendor releases and filing support tickets.
And you save something the spreadsheet cannot capture. A revenue system that matches your business instead of a business that contorts to match a vendor’s system.
“But What About Ecosystem and Integrations?”
I hear this one a lot. Usually from the person who manages the vendor relationship and has built a career around it.
The argument goes like this: the CRM is not just a database. It is an ecosystem. AppExchange. Marketplace. Thousands of integrations. You cannot replicate that.
That is true. You cannot replicate it. You also do not need to.
Count the AppExchange apps you actually use. Not the ones you installed and forgot about. Not the ones someone evaluated two years ago. The ones your team uses this week.
It is probably four. Maybe six.
An enrichment tool. A call recording integration. A document generation plugin. Maybe an e-signature connector. Those are API integrations. Your engineering team can build those connections in days, not months. And unlike the marketplace versions, yours will work exactly the way your business needs them to work instead of the way someone else’s product manager designed them to work.
The ecosystem argument was strong when building integrations was expensive. It is not expensive anymore. An engineer with agentic tooling can build an API integration in an afternoon that would have taken a team two weeks in 2022.
The Courage Problem
Here is why most companies will not do this even though the math is obvious.
It is not a technology problem. It is not even a budget problem. It is a courage problem.
Someone signed that CRM contract. Probably someone senior. Probably someone still in the room. And admitting that a decision that felt strategic five years ago is now a tax on the business — that is a hard conversation.
The board is used to seeing that logo in the technology stack. The CFO has the renewal in the budget forecast. The sales leadership team has trained an entire generation of reps on the existing workflow. Procurement has a relationship with the account team.
None of those are reasons to keep paying. They are reasons nobody wants to be the first person to say stop.
That is inertia dressed up as strategy.
I have watched this movie before. On-prem infrastructure that nobody would kill because the VP who approved the data center was still in the room. Monolithic codebases that nobody would decompose because the migration felt scarier than the pain. Manual QA that nobody would automate because the test team’s jobs were on the line. The pattern is always the same. The incumbent is expensive and limiting. The alternative is cheaper and better. Everyone knows it. Nobody wants to be the one who says it out loud. Someone finally does. Eighteen months later, the entire industry has moved and the holdouts are scrambling.
The CRM is at that inflection point right now.
Start With One Wedge
I am not telling you to rip out your CRM on Monday. That would be reckless, and I am not in the business of reckless advice.
I am telling you to pick one wedge. The revenue workflow that matters most. Pipeline inspection. Renewal management. Executive forecast review. Board reporting. Pick one.
Give four engineers one clean mandate. Build the version your revenue leaders would choose if they were not constrained by the incumbent system. Build it on your data. Build it around your sales motion. Make it smart — let the AI layer do the things the CRM structurally cannot.
You will have a working prototype in a week. You will have real sellers using a real tool in four weeks. You will have evidence in eight weeks.
Not vendor mythology. Not a reference call with another company whose business looks nothing like yours. Not a demo from a sales engineer who rehearsed it forty times. Evidence. From your company. With your data. Against your workflows.
That is when the conversation changes. Because once your VP of Sales uses a tool that actually fits how your company sells — one that updates itself, flags risk, writes its own summaries, and does not require a two-day sprint to produce a board package — they are not going back to the old system.
Nobody goes back.
The Real Question
This is not a software question. It is a leadership question.
Are you willing to say out loud that a category of software your company treated as sacred infrastructure for fifteen years is now just another build-vs-buy decision — and the math has flipped?
Because your VP of Sales already knows. Your RevOps team already knows. Your sellers already know — they told you every time they exported data into a spreadsheet to do the work the CRM was supposed to handle.
The only people who do not know are the ones who have not looked at the economics. Or the ones who looked and did not want to have the conversation.
Four engineers. Twelve weeks. Your data. Your sales motion. A system that gets smarter every month instead of more expensive.
Next Monday at 9:14, your VP of Sales will open something. Right now it is a spreadsheet — because the system you pay $2.2 million a year for is not where work happens.
You can keep paying rent on the building they already left. Or you can build the one they would actually use.
