TL;DR
Tip
Most companies are held together by more “good enough” systems than anyone likes to admit.
That is not necessarily a criticism. You make the best decision you can with the time, information, and resources available. You create a field in the CRM. You add a workflow. You build a report. You solve the immediate problem and move on to the next one.
The process is not perfect, but it works.
We have done the same thing. Over the years, we built a meaningful amount of automation in Zoho CRM , including workflows to capture sales-cycle dates automatically . Deals moved through the pipeline. Renewals were created. Account values were updated, and we captured much of the information we needed to operate the business.
But “good enough to operate” is not always the same as “good enough to analyze.”
That difference becomes very clear when you start asking questions such as:
Those are not obscure finance questions. They are questions operators should be able to answer, and they are the kinds of questions lenders, investors, and potential acquirers will eventually ask.
Investopedia defines KPIs as quantifiable measures used to evaluate progress toward specific business objectives. The important word is not quantifiable. It is evaluate. A number is only useful if you can trust what it represents.
We had plenty of data.
We had years of Deals, Accounts, renewals, amounts, dates, statuses, and automation. The problem was that the system had evolved one practical decision at a time. Fields created for one purpose were later used for another. A total deal amount might sit beside monthly recurring revenue. A renewal might be identifiable by its name and date but not linked explicitly to the deal that came before it. A closed-lost renewal might imply churn even when the commercial outcome needed more explanation.
None of those choices stopped the company from operating.
Together, however, they made historical analysis harder than it should have been. A report could return a number without giving us enough confidence that it represented the business concept named at the top of the column.
This is where a lot of KPI work goes wrong. The formula is usually the easy part. The hard part is determining whether the underlying events were captured consistently enough for the formula to mean anything.
For example, net revenue retention is not simply “current revenue divided by old revenue.” It requires a defined group of customers, a consistent period, and a clear treatment of churn, contraction, and expansion. The SaaS Metrics Standards Board’s NRR standard specifically compares revenue from the same customer cohort over time and excludes revenue from customers added during the period.
If the CRM cannot distinguish those movements reliably, the dashboard may look polished while the number underneath it remains questionable.
I have used ChatGPT for plenty of familiar things: research, drafting, analysis, formulas, and troubleshooting. Those uses can save time, but connecting ChatGPT to Zoho CRM through Zoho CRM’s MCP tools changed the kind of work we could do.
MCP stands for Model Context Protocol. The useful part is simple: instead of describing the CRM from memory, copying screenshots, or exporting one report at a time, I could give ChatGPT controlled access to inspect the actual system.
That meant we could examine:
That compressed the investigation loop dramatically.
Before, I might have opened a workflow, followed it to a function, copied part of the code, searched for the fields it updated, exported sample records, and tried to hold the whole chain in my head. Then I would have needed to document the issue, decide on a better model, implement it, and test whether I had broken something else.
Now, I could work through that chain in one continuous conversation. ChatGPT could inspect the actual configuration, help trace the dependencies, compare the logic to the business outcome I wanted, and turn the findings into a concrete implementation plan.
It did not remove the need for judgment. It removed an enormous amount of friction between the question and the evidence.
This work would have taken me weeks or months of fragmented effort in the past. More realistically, some of it would have stayed on my list indefinitely.
That is the honest value of the efficiency gain. AI did not merely help me complete work faster. It made important work economically possible.
There is a long list of systems work that experienced operators know they should do:
None of it is flashy. Most of it loses to a customer issue, a product priority, or a sales opportunity when the calendar gets tight.
Combining ChatGPT’s ability to reason across a complicated system with Zoho’s ability to expose the live configuration changed that tradeoff. The amount of uninterrupted time required became much smaller, so work that had been important-but-deferred could finally move.
One of the most important conclusions from this work was that historical KPIs cannot depend entirely on mutable Account totals.
If a field always shows today’s recurring revenue, it cannot also tell you what recurring revenue was at the end of last March. Once the value changes, the prior state is gone unless it was preserved somewhere else. It is the same reason I have long argued that teams should date-stamp important CRM changes : current state and historical movement answer different questions.
The better design is an immutable monthly snapshot with a reconciliation that explains how the business moved from one month to the next:
Beginning MRR + New MRR + Expansion MRR + Reactivation MRR - Contraction MRR - Churned MRR = Ending MRR
That foundation will allow us to calculate and explain metrics including:
The SaaS Metrics Standards Board defines ARR as annualized subscription recurring revenue and notes that one-time fees and professional services should not be included. That sounds simple until your historical Deal records contain both recurring and nonrecurring amounts. Then it becomes a data-model and process problem, not a multiplication problem.
This is also why our work has gone beyond correcting old data. We are strengthening what the CRM captures going forward:
The result should be more investor-ready reporting, but it should also make us better operators. We will be able to see where revenue is growing, where it is leaking, and which customer or product patterns deserve attention.
There is a temptation to use AI to create a confident-looking answer and call the job finished.
That would be the worst possible use of these tools.
Investor-ready reporting is not about making the numbers look impressive. It is about making the definitions explicit, the calculation repeatable, and the limitations visible. Public companies do this as well. In an SEC filing, Bentley Systems explains both how it defines ARR and how it calculates dollar-based net retention , while also warning that non-GAAP measures can differ between companies.
That disclosure matters. Two businesses can use the same label and calculate it differently.
Our historical data still requires backfilling, reconciliation, exception review, and testing before every KPI should be treated as final. ChatGPT can help identify inconsistencies and accelerate the work, but it cannot invent evidence that was never captured. Where history is ambiguous, we need to say so.
That is not a weakness in the process. That honesty is part of what makes the eventual numbers more credible.
ChatGPT could inspect a renewal workflow. It could see that one field was empty or that a function used total Deal Amount where recurring revenue would have been more appropriate.
It could not decide what one subscription means for our company without my input.
We have multiple products. Some customers buy one, some buy both, and the products may renew together or independently. That commercial reality determines whether two Deals belong to the same subscription chain. No generic model can infer the right policy without someone who understands the contracts, the products, and how customers actually buy.
That division of labor is important:
The best results came from combining both.
I started this work because I wanted better historical KPIs and stronger investor and lender reporting.
What I found was a larger opportunity.
We are not just building a way to answer diligence questions. We are building a more accurate model of our own company. The same data that helps explain the business to an outside party also helps us run it:
Those questions should not require a heroic spreadsheet project every time someone asks them.
The combination of ChatGPT and Zoho’s MCP tools has helped us move from reports that were useful in the moment toward a system that preserves what happened and explains why.
The efficiency gain is substantial. But the real payoff is better understanding.
Our old processes were not failures. They were good enough to get us here. The opportunity now is to use better tools to build the system we will need for where we are going next.
Keep reading
We leaned into BrokerPro AI because practical AI belongs inside the work itself, where it can remove drag without asking teams to change how they operate.
Why I’m writing again — and how this site connects my work, expertise, and role across Infinity Software Solutions, BrokerPro, BrokerOS, and the changing world of AI-driven discovery.
People buy when the problem feels emotionally and operationally real, so strong selling ties your offer to what actually matters to them.
This website uses cookies to improve functionality and user experience. By continuing to use this site, you acknowledge our use of cookies. See our privacy policy .