All insights

Digital Transformation & Automation

Fix the process before you buy the software

A new CRM or operations tool rarely fixes a messy process. It records the mess faster. Here is a practical sequence for designing the process first, then choosing and configuring the software.

GainWells · 16 November 2026 · 6 min read

A familiar story. The sales pipeline is hard to read, follow-ups slip, and nobody trusts the forecast. Someone proposes a new CRM. Six months and a significant bill later, the same problems are back, now stored in a more expensive system.

The software was not the issue. The process was never clear enough to put into software.

This article sets out why that happens, how to spot it early, and a simple sequence to follow before you sign a contract.

Software records a process. It does not create one

A CRM, an ERP module or a workflow tool does what it is configured to do. It needs answers to basic questions:

  • What are the stages of a sale, and what must be true to move from one to the next?
  • Who owns each step?
  • What happens when something is late, missing or disputed?
  • What information is mandatory, and who is responsible for keeping it accurate?

If your team cannot answer these consistently today, the tool will force someone to guess. Usually the vendor's default settings or the most confident person in the room decides. Neither is a good basis for how your business runs.

Signs you are buying software to avoid a process decision

Some patterns come up repeatedly:

  • Different people describe the same process differently. Ask three team members how a lead becomes a customer and compare the answers.
  • The brief is a feature list. Requirements read like a vendor comparison sheet, not a description of how work should flow.
  • Spreadsheets run the real process. The official system exists, but people keep personal trackers because they do not trust it.
  • Exceptions are the norm. Every deal or order is described as special, so there is no standard path to build.
  • No one owns the outcome. The project belongs to IT or to a single enthusiastic manager, not to the person accountable for the result.

None of these means you should not buy software. They mean the work that comes first has not been done.

A practical sequence

You do not need a long consulting exercise. You need a short, honest piece of design work, done with the people who actually do the job.

1. Write down the process as it really runs

Not the version in the handbook. Map what happens from trigger to outcome: a lead arrives, an order is placed, a support request is raised. Use a whiteboard or a shared document. Include the workarounds, the spreadsheets and the messages sent on chat apps.

The aim is to see the work, not to judge it.

2. Find where work stalls, repeats or disappears

Mark the points where:

  • information is re-entered by hand
  • work waits for approval with no deadline
  • ownership is unclear
  • errors are found late
  • customers have to chase

These are usually where software can help. They are also usually where a rule or a role change would help more cheaply.

3. Design the process you want

Decide the stages, the owners, the entry and exit criteria for each step, and the handling of common exceptions. Keep it as simple as the business allows. A process with six clear stages that everyone follows beats one with fifteen that people skip.

Test it on paper against real recent cases. Pick a handful of past deals or orders, including awkward ones, and walk them through. Gaps show up quickly.

4. Define what the software needs to do

Only now write requirements. They should come directly from the designed process:

  • the stages and fields you need
  • the rules for moving between stages
  • the reports leaders will actually use
  • the integrations that remove re-entry, such as accounting, email or billing
  • who needs access and at what level

This turns a vague demo into a clear test. Ask each vendor to show your process, using your cases, not their sample data.

5. Pilot small before you roll out

Run the new process and tool with one team for a few weeks. Watch where people hesitate or work around it. Adjust the process as often as the configuration. Then extend it.

An illustration: the CRM

Take a hypothetical services business that wants a CRM to improve forecasting.

Before buying, it agrees that a deal only counts as "proposal sent" when a scoped quote has gone to a named decision-maker. It agrees that a deal with no recorded next step for a set period is flagged for review. It decides that one person, the sales lead, owns pipeline data quality.

These are management decisions, not software settings. Once made, the CRM configuration is straightforward, and the forecast improves because the stages mean something. Without them, the same CRM would hold stages that each salesperson interprets differently, and the forecast would be no more reliable than before.

The same logic applies to order handling, onboarding, invoicing, procurement and support. In each case the value comes from clear definitions and ownership, with software making them easier to follow and easier to see.

When to automate, and when not to

Automation is tempting because it feels like progress. A good rule: automate a process only once it is stable and understood.

Automating an unclear process makes errors happen faster and at greater scale. It also makes them harder to find, because the work is now hidden inside rules few people remember writing.

Sensible candidates for early automation tend to be:

  • repetitive, rule-based tasks with few exceptions
  • data transfers between systems that people currently retype
  • reminders and escalations that depend on someone remembering
  • standard documents generated from existing data

Leave judgement-heavy steps with people until you understand them well. If you are considering AI tools for parts of the work, the same discipline applies. Our overview of AI enablement covers how to decide where it fits.

The costs of skipping this step

The costs are mostly indirect, which is why they are easy to underestimate:

  • Low adoption. People return to spreadsheets because the system does not match how work happens.
  • Poor data. Reports are built on fields filled in differently by each person, so leaders stop trusting them.
  • Expensive rework. Reconfiguring a live system, retraining staff and migrating data costs more than designing properly at the start.
  • Vendor dependence. Without a clear process, you rely on the vendor's view of how your business should run.
  • Change fatigue. A failed rollout makes the next, better-planned initiative harder to land.

Questions to ask before you sign

  • Can we describe the target process in one page, and do the people doing the work agree with it?
  • Who is accountable for the result, not just the implementation?
  • Which problems would a rule or role change fix without any software?
  • How will we know in three months whether this worked? What would we measure?
  • What is the plan for data quality once the system is live?

If these are hard to answer, that is useful information. It tells you where to spend the next few weeks.

What to do next

Before your next software decision, set aside a short working session with the people who run the process. Map it as it is, mark the stalls, and agree the version you want. Write requirements from that, and test vendors against it.

If you would like a second opinion on a planned CRM or operations project, or help designing the process behind it, see our work on digital transformation and automation or contact GainWells to discuss it.

What is your next important decision?

Define the case and path forward for a new venture, market entry, growth or transformation.