Before we automate our sales process, define what good looks like

Sales leaders we work with are under pressure to put AI to work. The board wants it, the reps are already experimenting with it, and every vendor demo promises a faster pipeline. So the natural first question becomes, “What can AI do for us?”

There is a better question to ask first, and it is older and less comfortable: have we actually defined what good looks like in this part of our sales process?

Give it a great process and we multiply our best selling. Give it an inconsistent one and we multiply that inconsistency just as quickly.

The real problem is sequence

It is easy to get the order backwards. We start with the technology, pick a tool, and then go looking for a process to point it at. Six months later the tool is generating activity, the dashboards look busy, and we still may not know whether the selling actually got better. The tool did its job. We just had not fully defined the job yet.

The fix is not more technology evaluation. It is process definition. Before a single dollar goes to software, we should be able to answer: what does a great version of this workflow look like at our company? What does a great discovery call cover? What does a great proposal contain? What does a great handoff to delivery include?

If the answer changes significantly by rep, that is a useful signal that we still have a standard to define. We may have talented individuals doing excellent work, but AI needs that work to be visible before it can help us repeat it.

What this looked like in practice

We watched this play out with one of our clients, a managed services provider that sells complex hosting and infrastructure work. Their CPQ system held hundreds of products and did its job on pricing. What it could not do was produce the polished, narrative proposal a customer actually wants to sign, so reps were assembling proposals by hand from the company’s contract library.

Every rep did it a little differently. The proposals were not doing the company’s work justice, and the obvious move was to buy proposal software.

We advised the opposite. Buying a tool at that moment would have automated the inconsistency. So we sat with the people closest to the work, the rep, the solutions lead, the sales-ops lead, and locked the proposal standard first. What must every proposal contain? Where does pricing come from? Which contract language is approved? What does the customer need to see before they sign?

Only after that standard existed did we build the AI. The agent takes the CPQ pricing, the deal notes, and the approved contract library, and produces a finished proposal. Pricing comes only from the quote. Terms come only from the approved library. Anything missing gets flagged instead of invented.

Within two weeks, the company walked into its national sales meeting with a working agent instead of a plan to buy one. The AI did not replace their sales judgment. It executed a standard their own team had defined.

Where P.R.E.D.I.C.T. comes in

This is where our P.R.E.D.I.C.T. framework comes in. It names seven focus areas of strong selling: Planning, Researching, Engaging, Discovering, Intersecting, Communicating, and Transitioning. Together, these areas give us a practical way to define what good looks like across a sales motion. AI can support every one of them. It cannot replace the discipline behind any of them.

An AI agent can research an account in minutes, but we still have to decide what our plan for that account should accomplish. It can draft a follow-up in our voice, but the trust that makes the follow-up land still has to be built. It can summarize a discovery call, but it cannot replace the judgment required to recognize what needs to be explored further.

The framework helps us define the standard. AI is the horsepower we apply to it.

Try this with one workflow

Here is an exercise worth an hour with our sales leadership teams. Pick one workflow, proposals, account planning, discovery prep, whatever creates the most friction, and run it through P.R.E.D.I.C.T.:

• Planning: what are we trying to accomplish?
• Researching: what information should inform the work?
• Engaging: how should we build trust and differentiation?
• Discovering: what must we learn from the customer?
• Intersecting: where will objections or alignment matter?
• Communicating: what value must we communicate?
• Transitioning: how should the relationship grow from here?

Answer those questions for that one workflow and we have something incredibly useful: a shared definition of the standard. From there, we can sort the work into three buckets. What should AI automate outright, such as assembly, formatting, or retrieval? What should AI support, such as drafts, research, or prep? And what should stay with human judgment, such as the relationship, the hard question, or the call on whether a deal is real?

That example makes the tradeoff clear. The harder part was not the software. It was the unglamorous work of agreeing on what good looks like. Once that exists, the AI decision gets easier, and often cheaper, because we finally have something meaningful to measure the tools against.

The part that stays ours

AI can scale what good looks like. It can take our best proposal, our best research motion, our best call prep, and help make that level of execution more consistent. What it cannot do is decide what good means at our company, for our customers, in our market.

Defining good is still our job. When we do that work first, we give AI something worth accelerating. That has been true across sales technology waves, and this one will not be different. It will just be faster.