sellers feel overwhelmed by the technology their job needs, and they are 45% less likely to hit quota.
7 in 10 already use AI tools, and more than 4 in 5 cite inaccuracy and poor data integration as obstacles.
Bring governed quoting to the places your sellers and buyers work.
CPQ means configure, price, quote. Headless CPQ separates those capabilities from the quote editor, so a chat, portal, or agent can use the pricing and approval controls you already trust. Start with one useful workflow and prove that its quotes hold up before adding more interfaces.
CPQ engines can already expose APIs. For example, Salesforce CPQ documents quote APIs. The common obstacle is an implementation that depends on editor scripts, copied price lists, or manual checks. Inventory those dependencies before deciding what needs to change.
Headless CPQ gives each interface access to governed actions such as configure product, price quote, and request approval. Reuse the existing engine where it works. Keep clear authority for each decision, even when pricing, agreements, and billing live in different systems.
“Quote Northwind for 200 seats with premium support and onboarding.”
Q-20418 · €86,000Quote Northwind: 200 seats, premium support, onboarding
quote.create({MCPseats: 200,addOns: ["support","onboarding"]})→ Q-20418 · €86,000 · pricedSellers, buyers, and agents all want to quote from somewhere new, and each of them needs the answer your deal desk would give.
sellers feel overwhelmed by the technology their job needs, and they are 45% less likely to hit quota.
7 in 10 already use AI tools, and more than 4 in 5 cite inaccuracy and poor data integration as obstacles.
of B2B buyers prefer buying without a rep.
69% got different answers from a supplier’s website and its reps, and inconsistent information is the top reason buyers switch suppliers.
of 400+ B2B pricing executives name data quality and integration as the top barriers to agentic AI in pricing.
AI agents per seller by 2028, Gartner predicts.
Forrester predicts 20% of B2B sellers will engage in agent-led quote negotiations in 2026.
Dated forecasts; adoption varies“If those systems are fragmented, the agents will scale the fragmentation.”
Gartner, July 2026
These sources describe pressure for easier buying and better-connected systems. The case for a CPQ investment still depends on your workflow and its current friction.
Look for recurring quote friction. Simple quoting that already works well may gain little from another interface.
Customer rates, amendments, and renewals need agreement context every time someone asks for a price.
Useful when manual checks regularly hold up a quote.See how it worksReps, partners, and customers request quotes in different places, with different permissions.
Useful when separate price lists or repeated entry create corrections.See how it worksTiers, usage, bundles, or proration make a copied price table unreliable.
Start where a governed calculation saves a recurring manual check.See how it worksThe same inputs, caller identity, agreement scope, and rule versions should produce the same result. Different seller and partner permissions still apply. Illustrative quote below.
Any rep surface
Same action
Same governed records
Same quote
* When a rep’s agent prepares the quote, it works under the rep’s identity and discount limit, and the rep confirms it before it goes out.
Illustrative pilot: one product, one agreement, one amendment, requested by a seller in chat.
The agent asks for the effective date and resolves the product and agreement by ID. It does not guess between two similarly named rate cards.
At €360 per seat per year with exactly six months remaining, 50 seats cost €9,000 before tax in this simplified example. The existing engine applies the actual agreement and proration rules under the seller’s identity.
Compare lines, totals, dates, and approvals with a trusted amendment. Wait for the document to finish and confirm it still matches the approved quote. The seller reviews and sends it.
Prove the failures too: missing dates, invalid quantities, ambiguous products, declined discounts, and pending approvals must produce a clear status. Measure preparation time, manual corrections, and approval accuracy against today’s process.
Aquiva’s work is to connect the interface to the governed capabilities, find dependencies that do not survive an API call, and verify the quote through to its document. Billing and provisioning can stay on their existing process during this pilot.
Whether a person in the portal or a procurement agent asks, the answer has to be the one your deal desk would give.
An existing customer adds seats, upgrades a plan, or checks out on the website. The price they see includes their contracted rates and proration to their renewal date, and it matches what their account team would have quoted.
A buyer’s procurement agent could connect through MCP (Model Context Protocol), an open protocol for calling tools. It requests a quote within the buyer’s authenticated account scope and can submit a counteroffer. The service returns the approved price, expiry, and any pending approval status, while keeping costs and margins private.
The example in Where it’s heading follows a buyer’s agent negotiating with your quoting agent, and shows where a person steps in.
These are design goals to verify in the selected workflow, rather than guaranteed outcomes of adding an agent.
Governed actions enforce discount limits, margin floors, and approvals under the caller’s identity. Each quote records the rules it used.
McKinsey: a 1% price increase lifts operating profit by about 8.7% on average, assuming volume holds.
Connected interfaces use the authoritative version. Saved quotes retain their version until a controlled recalculation.
See the channelsBilling receives agreed rates, commitments, and dates. Forecast consumption stays separate from actual billed usage; activation follows its own triggers.
Link the request, calculation, approvals, and agreement versions across their owning systems. An agent can explain the result without inventing a price.
Use the capabilities you already have. This pilot checklist does not require a full catalog migration or an automated quote-to-cash program.
Start
Pick a recurring task, such as an existing-customer amendment. Define who uses it, what it may change, and where a person confirms the result.
Measure preparation time, corrections, and approval accuracy today.Find the calculation engine, agreement data, and policy owners for that workflow. Inventory existing APIs and editor-dependent scripts. Reuse governed code; resolve only the gaps that block this scope.
Define which system owns each decision and how changes are synchronized.Define inputs, validation, permissions, approval states, and version behavior. Use the intended seller or customer identity. Add an API or agent tool where the selected interface needs it.
An administrator account is not proof that seller approvals work.Compare with trusted quotes, including missing data, ambiguous products, invalid requests, declined discounts, and pending approvals. Check the completed document after asynchronous work finishes.
A returned price alone does not make a quote ready to send.Pilot with the chosen users and confirmation step. Compare time, corrections, and approvals with the baseline. Add another channel or downstream automation when its value and controls justify it.
The first release can be internal or external; scope determines the controls.Release one proven workflow. Expand where it pays off.
The agent interprets a request, asks for missing details, and explains the result. Governed services calculate prices, validate inputs, enforce permissions, and determine approvals. An API contract defines the inputs and results; it is separate from the customer’s commercial agreement.
Whether a rep in the wizard, a partner in their portal, or a buyer’s agent asks, it is the same action with the same checks. Surfaces never write to the records directly.
A new channel or surface is a new caller of existing actions. Reuse pricing actions; verify the new caller’s access and behavior.
| Channel | Surfaces | What the API contract returns |
|---|---|---|
| Seller channels | CRM quoting wizard, voice, chat in Slack or Teams, rep agents | The rep’s identity and discount limit, drafts saved to the record |
| Partner channels | Partner portal, plus the chat, voice, and agent surfaces reps use | The partner’s identity and discount limit, the partner price list and margin, the end customer’s agreement, no view of the rules behind the price |
| Buyer channels | Customer portal, signed-in website checkout, in-product upgrade, marketplace listings | Contracted rates for known customers, proration, payment handoff, a readable held status |
| Agent channels | AI agent workspaces and procurement platforms, through MCP tools | Named MCP tools mapped to actions, per-agent limits, no cost or margin exposed, a trace on every call |
Select the capabilities your workflow needs. These examples are not prerequisites for every headless implementation.
Identify who owns products, agreements, assets, and pricing decisions. Retrieve the relevant versions for the scoped quote; they can remain in connected systems.
Price lists by region and currency, plus tiers.
One-time, recurring, usage-based, or a mix, plus trials.
What it includes, requires, and can’t be sold with.
The template and clauses legal approved.
The access it grants, with agreed start dates and payment conditions.
Every product, quantity, and end date, so new lines share the renewal date.
Signed terms that replace template clauses, for every subsidiary it covers.
Special rates, price holds, and capped uplifts, applied first.
How much of what they bought they use, with low usage flagged early.
Salesforce may own accounts, contracts, and orders while another service owns pricing or billing. Define the authority and synchronization behavior for each.
For the selected workflow, identify the applicable rules and their owners. Governed calculation code can remain authoritative.
A typical split, and yours will differ. Agents use rules and read records, and change quotes only through actions.
| What can change | Product | Pricing ops | Finance | Legal | Sales + partners | Agents |
|---|---|---|---|---|---|---|
| Products and versions | Owns | Reviews | Reviews | · | Reads | Reads |
| Price rules and tiers | Proposes | Owns | Approves | · | Uses | Uses |
| Customer-specific pricing | · | Reviews | Approves | · | Proposes | Uses |
| Promo and referral codes | · | Owns | Approves | · | Uses | Uses |
| Discount limits by role | · | Proposes | Owns | · | Uses | Uses |
| Contract templates and clauses | Reviews | · | · | Owns | Uses | Uses |
| Master service agreements | · | · | Reviews | Owns | Proposes | Reads |
| Approval chains | · | Proposes | Owns | Owns | Reads | Reads |
Gartner predicts that by 2028, 90% of B2B buying will be AI agent intermediated, and the protocols for it already exist: A2A lets agents from different companies talk, and MCP lets them use a company’s tools. This is a forecast. The example below explores a possible extension, once its commercial authority and controls are defined.
A buyer’s agent can call your MCP tools directly, or talk to your own agent, which calls the same tools. Either way the same actions and limits apply: your agent requests quotes through governed services within the agreed limits, a person approves anything above them, and the buyer signs. The numbers are illustrative.
Requested byBuyer’s procurement agent
TermTwo years
Valid14 days
| Line | Billing | Price |
|---|---|---|
| Platform, 600 seats, 500+ tier | Annual | €192,000 |
| Premium support, MSA rate | Annual | €24,000 |
| Phased onboarding, half price with LAUNCH26 | One-time | €3,000 |
“…headless architectures will establish a significant competitive moat.”
Gartner, October 2025
Choose one workflow first. Answer for that scope and the capabilities you already have; this is a suggested next check, not an enterprise maturity score.
Have you chosen one quoting workflow, its users, and a baseline for time and corrections?
Can the needed actions run under the intended caller’s identity, enforcing permissions and returning clear approval states?
Have you checked trusted quote results, missing and invalid inputs, declined or pending approvals, and the completed document?
Is the selected interface ready for a small pilot, with a confirmation step and a plan to measure time, corrections, and approvals?
Answer all five to see the next capability to check.
0 of 5 answeredTell Aquiva which workflow you want to improve, who uses it, and what gets in the way. You can include the checklist answers or write your own note.