Salesforce Revenue Cloud will price every quote. Should it decide them too?

Every Revenue Cloud implementation automates pricing. That’s what the platform is for. So the debate about whether to automate pricing is the wrong debate. The question that matters now is what the system should be allowed to decide on its own, and what it should take to earn that.

AI agents can already build quotes, recommend discounts and route approvals. Most businesses will switch those capabilities on before they’re ready. They won’t get caught out by the technology. They’ll get caught out by a pricing policy that only ever existed in people’s heads.

Revenue Cloud manages the whole revenue lifecycle: what you sell, how it’s priced, quoted and approved, and how the contract, asset and invoice stay consistent as the relationship changes over time.

What gets overlooked is that it’s usually the first time a business has to write its pricing logic down. Years of judgement calls, regional habits and customer-specific deals have to become rules a system can follow. Most companies discover during implementation that their "pricing policy" is really just years of unwritten habits and gut-feel decisions.

That makes a Revenue Cloud programme a pricing strategy exercise as much as a technology project. The businesses that treat it that way get far more out of it.

Most of the confusion comes from treating pricing automation as one thing. It’s three, and the risk sits in only one of them.

1. Calculation. The engine works out the price from the list price, attributes, volume, term and contracted rates. It’s always automated, and it should be. No rep should calculate a price by hand.
2. Governance. Discount guardrails, approval thresholds and automatic approval within agreed bands. People write the rules and the system enforces them. This has been standard practice for a decade.
3. Decisioning. The system decides for itself: it recommends the discount, builds the quote, grants the approval. This is new, and it changes who is accountable for the price a customer sees.

Calculation and governance carry out decisions people have already made. Decisioning makes new ones. A business can be fully mature in the first two and nowhere near ready for the third.

When automated pricing goes wrong, the instinct is to blame the AI. In our experience the AI is rarely the cause. It just works faster, more consistently and at greater scale on whatever it was given.

Manual delays are obvious to everyone. But a fast, confident, wrongly-priced quote looks just like a correct one and sails straight through without a second glance.

The failures follow a familiar pattern:

• Renewals are priced from the wrong base, because migrated subscriptions don’t match what the customer actually signed.
• Invalid product combinations quoted without complaint, because a dependency was never captured as a rule.
• Approval thresholds that quietly stop working, because they were set against list prices that have since moved.
• Exceptions that become policy by accident, because a regional carve-out or legacy deal was codified without anyone deciding it should be.

None of these is an AI failure. They are pricing decisions nobody made explicitly, now carried out at machine speed.

Most conversations about AI readiness focus on data quality in general terms. In pricing, readiness comes down to four specific things:

1. The catalogue. What you actually sell, how it’s packaged and what can be combined with what.
2. The rate card. The prices themselves, including the tiers, attribute-based prices and contracted rates that make up the real price list.
3. The installed base. What each customer has today and on what terms. Every renewal and amendment depends on it.
4. The policy. Who can approve what, and why, including the exceptions so far handled by judgement.

The catalogue matters most, because everything else is built on it. A sprawling catalogue full of one-off variants, regional duplicates and bundles nobody sells can’t be governed, let alone handed to an agent. Across the Revenue Cloud programmes we’ve delivered, the effort and risk track how complex the catalogue is, not how big it is. Where simplification was left late, it set the go-live date.

Revenue Cloud rewards decisions made early and punishes decisions made late. Some platform choices are hard or impossible to reverse once data starts flowing. A pricing error on one quote is easy to fix. The same error repeated across thousands of quotes, contracts and invoices becomes a remediation programme.

So design time is where the value is. The questions settled before the build (what the catalogue looks like, how prices are structured, where judgement sits) decide whether the platform can ever be trusted to act on its own.

Product simplification is the step most implementations skip, and the one everything else depends on. Carrying years of legacy complexity into a new platform doesn’t modernise it. It gives the same problems better formatting.

Simplification means collecting everything currently sold, quoted and billed, retiring what doesn’t earn its place, collapsing variants into configurable products and building one catalogue that sales, finance and operations all work from.

The result is simpler for customers to buy from and simpler for reps to sell from. It’s also the only foundation an AI agent can safely act on. Simplification isn’t preparation for automation. It’s the strategic work that makes automation worth having.

AI decisioning shouldn’t be a feature you switch on at go-live. It’s a licence the system earns through evidence. We recommend four stages:

1. Baseline. Measure how quoting works today: turnaround time, escalation rates, rework. Without a baseline there’s no way to prove the system is improving, or that it’s ready for more responsibility.
    
2. Simplify and configure. Rationalise the catalogue, settle the rate card and policy, and build them into the platform. Calculation and governance are fully automated from here.
    
3. Assist. The system recommends and a person decides. Every accepted, overridden or corrected recommendation is evidence of whether the underlying logic holds.
    
4. Act within limits. Hand decisions to agents only for the deal types where people have consistently stopped overriding them. Send everything else to a person, and revisit the limits whenever products, prices or policy change.

The override rate is the most useful metric in this whole process. When people consistently stop correcting the system, it has earned the right to decide. Until then it hasn’t, however capable the technology looks.

Revenue Cloud should price every quote automatically from day one. Whether it should decide them is a different question, and the answer depends on how clearly a business has defined what it sells, what it charges and who gets to say yes.

The businesses that get the most from AI in pricing won’t be the ones that adopt it first. They’ll be the ones whose catalogue, pricing and policy are clear enough for a machine to follow.

If you’re planning a Revenue Cloud implementation or a move from Salesforce CPQ, and want an independent view of whether your catalogue and pricing are ready for what comes next, contact us or email hello@solutionjunkies.co.uk.