Sellers
1 in 2sellers 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.
Aquiva POV
Headless CPQ
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.
Sellers and buyersAny surface
- Sales and partner repsVoice, chat, wizard, own agent
- CustomersPortal, website, in-product
- Buyer agentsProcurement tools
Governed capabilitiesPricing and approvals
- Products + versions
- Price rules
- Customers + MSAs
- Contract terms
- Approval policies
Connected as neededApproved terms
- Contract
- Billing
- Provisioning
- Renewal
What it is
What changes when CPQ goes headless.
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.
When channels work around the engineEditor dependencies leave other channels waiting.
- Customer portalcopied price listOut of sync
- Partner portalemails the deal deskOut of sync
- Websiteits own price tableOut of sync
- Buyer agentswait for a repWaiting
Voice
“Quote Northwind for 200 seats with premium support and onboarding.”
Q-20418 · €86,000 Chat
Quote Northwind: 200 seats, premium support, onboarding
Q-20418€86,000Open quote
Quoting wizard
Step 2 of 4 · Products
- Platform, 200 seats€72,000
- Premium support€8,000
- Onboarding€6,000
- Analytics add-on
Q-20418€86,000
BackNext
Customer portal
Platform · 200 seats−200+
- Premium support
- Onboarding
€86,000 Buy
Buyer agent
quote.create({MCPseats: 200,addOns: ["support","onboarding"]})→ Q-20418 · €86,000 · pricedWhy now
Every surface is becoming a place to quote.
Sellers, buyers, and agents all want to quote from somewhere new, and each of them needs the answer your deal desk would give.
Buyers
67%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.
Data
Over halfof 400+ B2B pricing executives name data quality and integration as the top barriers to agentic AI in pricing.
Agents
10 : 1AI 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.
Who it matters to
Where the investment can earn its place.
Look for recurring quote friction. Simple quoting that already works well may gain little from another interface.
Negotiated agreements
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 worksSeveral quoting channels
Reps, partners, and customers request quotes in different places, with different permissions.
Useful when separate price lists or repeated entry create corrections.See how it worksComplex pricing
Tiers, usage, bundles, or proration make a copied price table unreliable.
Start where a governed calculation saves a recurring manual check.See how it worksSeller experience
Different interfaces. The same governed calculation.
The 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
- Voice
- Chat
- Quoting wizard
- Agent-assisted*
- Rep’s own agent*
Same action
Price quoteNorthwind · 200 seatsas a direct rep, EU
Same governed records
- Product catalog
- Platform (per seat, annual)
- Premium support (annual)
- Onboarding (one-time)
- Price rules
- Direct EU price list
- 200-seat tier
- 5% rep discount limit
- Customer record
- Northwind MSA with a contracted support rate
Same quote
Q-20418 · Northwind Logistics
- Platform, 200 seats€72,000price rules
- Premium support€8,000customer record
- Onboarding€6,000price rules
- Approvalsnoneprice rules
- Total€86,000
* 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.
A first workflow
An existing customer asks to add 50 seats.
Illustrative pilot: one product, one agreement, one amendment, requested by a seller in chat.
“Add 50 seats to Northwind’s subscription.”
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.
Contracted rate, proration, and permissions.
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.
A draft is checked before it is sent.
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.
Buyer experience
The buyer may be a person or an agent.
Whether a person in the portal or a procurement agent asks, the answer has to be the one your deal desk would give.
A customer prices a change in the portal or on your website.
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.
- Customer portal
- Website checkout
- In-product upgrade
The buyer sends an AI agent to get the quote.
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.
- Connects through MCP
- Same price as the portal
- Approval status and expiry
- No cost or margin exposed
The example in Where it’s heading follows a buyer’s agent negotiating with your quoting agent, and shows where a person steps in.
Benefits
What governed quoting makes possible.
These are design goals to verify in the selected workflow, rather than guaranteed outcomes of adding an agent.
Every price follows the same policy, and can be explained.
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.
Publish a rule once, with an effective date.
Connected interfaces use the authoritative version. Saved quotes retain their version until a controlled recalculation.
See the channelsCarry approved terms into the next system.
Billing receives agreed rates, commitments, and dates. Forecast consumption stays separate from actual billed usage; activation follows its own triggers.
A quote you can trace and explain.
Link the request, calculation, approvals, and agreement versions across their owning systems. An agent can explain the result without inventing a price.
How to get there
A practical path for the first workflow.
Use the capabilities you already have. This pilot checklist does not require a full catalog migration or an automated quote-to-cash program.
Start
Choose one quoting workflow.
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.Map the authority you already have.
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.Expose actions with the right identity.
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.Prove the quote before sending it.
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.Release the interface, then measure.
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.
Architecture
Interfaces request. Governed services decide.
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.
- SurfacesWhere quotes are requestedClick-based
- Quoting wizard
- Partner portal
- Customer portal
- Website checkout
Prompt-based- Voice and chat
- AI agent workspaces
- Agentic layerOnly where the request is a promptClick-basedAPIsPrompt-based
- Reads intent from a prompt
- Asks for missing details
- Calls governed tools
- Explains results and routes exceptions
- API contractWhat each caller may do
- APIs
- MCP tools
- Caller identity
- Quote status
- Trace
- ActionsEach business action, defined once
- Configure product
- Price quote
- Apply discount
- Apply promo code
- Add seats
- Request approval
- Generate document
- Accept quote
- Start trial
- Renew contract
- Authoritative systemsOwned data and governed calculation
- Products and versions
- Price rules
- Discount limits by role
- Customers and MSAs
- Templates and clauses
- Approval chains
- Quotes, contracts, orders
- Connected systemsOptional handoffs, with explicit triggers
- E-signature
- Billing
- Provisioning
- Usage
- ERP
One action, defined once
Apply discount
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.
- Inputs
- The quote, the discount in percent or amount, line or quote scope, and the caller.
- Checks
- The caller’s discount limit, the margin floor for each product, the stacking rule with codes and contracted pricing, and the account scope.
- Result
- The request is applied, declined with a reason, or held for the required approver, with an expiry where applicable.
- Trace
- Who asked, through which channel and surface, which checks ran, and what changed.
Channels, and the surfaces they quote from
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 |
Technical checklist: data, rules, permissions, and ownership
Requirements
What the solution has to do.
Select the capabilities your workflow needs. These examples are not prerequisites for every headless implementation.
Identify the owning systems
Identify who owns products, agreements, assets, and pricing decisions. Retrieve the relevant versions for the scoped quote; they can remain in connected systems.
Product recordPlatform subscription
- Version
- 3, effective 1 January 2026
- Product line
- Core platform
- Status
- Active, version 2 retired for new sales
- Price rules
Price lists by region and currency, plus tiers.
- Billing model
One-time, recurring, usage-based, or a mix, plus trials.
- Bundle rules
What it includes, requires, and can’t be sold with.
- Contract terms
The template and clauses legal approved.
- Provisioning
The access it grants, with agreed start dates and payment conditions.
Customer recordNorthwind Logistics
- Parent
- Northwind Holdings
- Agreement
- MSA signed July 2024, covers all subsidiaries
- Status
- Customer since 2023, three active subscriptions
- What they own
Every product, quantity, and end date, so new lines share the renewal date.
- Master agreement
Signed terms that replace template clauses, for every subsidiary it covers.
- Contracted pricing
Special rates, price holds, and capped uplifts, applied first.
- Usage
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.
What the price rules have to cover
For the selected workflow, identify the applicable rules and their owners. Governed calculation code can remain authoritative.
- Geography and currency
- A price list per region and currency.
- Quantity and tiers
- A rate per tier, for all units or only those in the tier.
- Customer-specific pricing
- Holds and capped uplifts from the agreement, applied first.
- Bundles
- Required, optional, and excluded parts, priced together or apart.
- Promo and referral codes
- An owner, dates, eligible products, a limit, and a stacking rule.
- Discounts
- Percent or amount, per line or quote, in a fixed order.
- Proration
- Mid-term additions charged to the contract end date.
- Billing models
- One-time, monthly, annual, usage-based, or a mix.
- Trials
- A zero-price line with an end date that converts or expires.
Functional
- Pricing data and calculation code have owners, versions, and effective dates where applicable.
- The scoped workflow reads the relevant customer assets, agreement coverage, and contracted rates.
- Approval routing preserves policy for the authenticated caller: role, discount, margin, product, term, and region. Account hierarchy determines agreement coverage and access.
- Documents come from legal-approved templates and go out for e-signature.
- If downstream handoffs are in scope, transfer approved terms and reconcile status; start dates and payment conditions control billing and provisioning.
- Usage-priced quotes distinguish contracted rates, minimum commitments, forecast consumption, and actual billed usage.
Non-functional
- Each exposed action has validated inputs, a documented API contract, and explicit success, pending, and declined states.
- Each caller has an authenticated identity and permissions for the requested actions.
- The same inputs, identity, agreement scope, and rule versions produce the same result on every interface, with a trace.
- Changes that need approval get it before they publish.
- External agents never see cost or margin.
Governance: who may change what
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 |
Explore a future scenario: agents negotiating a quote
Where it’s heading
Two agents negotiate. A person signs off on the exceptions.
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.
Final quote · illustrativeQ-20431 · Northwind Logistics
Sent for signatureRequested 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 |
- Subtotal
- €219,000
- Deal desk discount, 10% on a two-year term
- −€21,900
- Year one
- €197,100
- Account executive600 seats
- Deal desk10% on a two-year term
“…headless architectures will establish a significant competitive moat.”
Gartner, October 2025
Self-assessment
Check one workflow, in five questions.
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?
Your next check
Answer all five to see the next capability to check.
0 of 5 answeredDiscuss your workflow
Choose the first quoting task to improve.
Tell 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.
- The quoting task and interface you have in mind.
- Your current engine, agreements, and approval process.
- The result you would use to judge a first pilot.