FAQ

Procurement teams usually have two options: personally negotiate a few high-value deals, or leave the long tail of smaller contracts and renewals unmanaged. Autonomous negotiation covers the spend that would otherwise go untouched, so your team can focus on the deals that need a person.

Three agents on one negotiation platform, each built for a specific job and tailored to your data, tone of voice and guardrails:

• Savings Campaign Agent: asks hundreds of existing suppliers for savings, better payment terms or acceptance of your standard terms, all at once. Suits tail spend.
• Quote Challenge Agent: improves a single supplier’s quote or pushes back on a renewal price increase. Suits services quotes, renewals and software licenses.
• Competitive RFQ Agent: gets the best result from 2–5 competing quotes for a defined scope, such as logistics or services.

A Custom Agent add-on handles related tasks such as simple NDAs, policy confirmations and document checks.

Most agents go live in 4–8 weeks, covering objectives, build, testing and launch. You can start with no integration, so you prove value before you scale. Results start as soon as the agent is live, and the first agreements typically arrive within days.

Yes. Most clients start with a pilot that needs no integration. We agree the scope with you, with a view to a longer-term partnership. Get in touch to talk through the right approach for your situation.

A pilot usually runs over three months:

1. We work with you to design the agent and gather the information it needs.
2. We build the agent, then upload your data, test it and optimize it (roughly month one).
3. The agent runs the negotiations with your suppliers (roughly month two).
4. We deliver the agreed terms as a spreadsheet, with a full transcript of every negotiation.

You provide the supplier or contract data, typically as a spreadsheet. The timeline includes some contingency. A pilot needs more than 100 negotiations, so you have enough data points to judge savings and ROI properly.

Afterwards, we work with you on integrations and features so your team can run the agent in your own environment.

Both. You define the use case and we build a Negotiation Blueprint for it: the limits, targets, terms and tone the agent works within. Once the agent is built, your team runs the negotiations, uploading a supplier list or the quote or contract up for renewal and following progress on the dashboard.

Three practical things matter more than technical sophistication: reasonably clean supplier and contract data, a category with enough volume to make results meaningful, and someone who can define the negotiation parameters and sign off the guardrails.

No. Better data improves results, but it isn’t a prerequisite. You can start with a spreadsheet of suppliers or a handful of quotes, with no integration. Waiting has a cost: each quarter, organizations negotiating their tail spend systematically pull further ahead on realized savings and on how ready their team and data are to scale.

Usually one. Team size and volume don’t drive the number of agents: one agent can support a team of any size, and there’s no volume limit per agent. You need a second agent only for a different type of negotiation. For example, one-to-many daily reverse auctions with logistics suppliers differ from one-to-one savings negotiations on indirect services in structure, tone and scope (price only versus price plus payment terms, start dates and legal terms).

It depends on how well defined the negotiation is. Fully autonomous agents suit negotiations with clear, repeatable parameters, such as price, payment terms or rebates within set ranges, at volume. A co-pilot suits negotiations that are less certain or run over weeks: the agent handles what it can, then drafts an email for a buyer to send or briefs them before a call.

No, it’s a hybrid. Generative AI lets the agent respond naturally to whatever a supplier says, using behavioral science and proven negotiation techniques. The commercial logic sits outside the LLM in a Negotiation Blueprint: what the agent can offer, the limits it can never cross, and how it opens and closes. 100% of offers stay inside the limits you set.

Not every negotiation ends in a new deal. Sometimes a supplier shows they already hold better terms than the ask, and the platform handles this without a category manager chasing it. Your team keeps full visibility and control throughout: every negotiation has a downloadable transcript and activity log, you can add human checkpoints at any decision point, and you can pause any negotiation at any time.

We test it against real outcomes, not just conversation quality. Different strategies (opening position, concession pace, how firmly to hold a line) run side by side on live supplier cohorts, and the ones that achieve better terms shape future negotiations.

Nibble is ISO/IEC 27001 certified and independently penetration tested. It runs on enterprise-grade cloud infrastructure, hosted in the UK by default, with regional hosting available. Data is encrypted in transit and at rest, access is role-based and logged, and single sign-on (OpenID, SAML) is supported.

No. Your data and your suppliers’ data are used only to deliver your service, never to train models for other customers.

Yes. Most clients start with a file upload (CSV, Excel or PDF) and no integration, so they prove savings before committing engineering time. Nibble then connects to platforms such as Coupa, SAP Ariba, Zip, Sievo and Microsoft Teams, via REST API or MCP (Model Context Protocol).

No. Nibble has completed 2M+ real negotiations, including more than 300,000 in a single weekend during one campaign. You can negotiate with many suppliers at once, with no volume limit.

It’s where your negotiation agent deals directly with an AI agent on the supplier’s side, rather than with a person, using open protocols such as MCP instead of a bespoke integration. It’s early, but worth planning for. Nibble already supports agent-to-agent communication via MCP.

Yes. Many clients start in English to keep the pilot simple, then add other languages once the approach is proven.

Only a web browser. Suppliers receive an email invitation with a secure link and negotiate in a chat window. There’s no login, software or account to set up.

We work with large organizations in airlines, FMCG, industrial manufacturing, chemicals and retail across the US and Europe, and with leading global management consultancies.

Where the negotiation has clear, comparable parameters and enough volume: tail spend and indirect categories with many similar suppliers, services and software renewals, and logistics or freight with repeatable RFQs.

Clients typically see price improvements of 7–10% and payment terms improved by 30+ days on average, on spend that usually goes unchallenged. Results vary by category, supplier base and campaign design, so we estimate the expected ROI with you before starting.

Against each supplier’s prior-year spend or pricing. Savings and term improvements are tracked against that baseline as standard, and every agreement is backed by a full transcript.

Add the savings and cash-flow benefit (price reductions, rebates, better payment terms) to the value of the negotiation time your team gets back, then subtract the cost of the agent. Payment-terms campaigns are usually the easiest to model first, because the cash release is concrete and immediate.

It depends on your use cases and the scale of rollout. Get in touch to discuss options that suit how you want to work.

Scope the pilot to need minimal sign-off. A payment-terms campaign, for example, usually involves only procurement and accounts payable, with no discount approval and no change to how other teams work. Scaling from a proven pilot is then simpler than rolling out everywhere at once.

Most do. Suppliers negotiate in free text, explain their cost pressures and propose alternatives, and the agent adapts. Engagement depends more on accurate contact details and well-timed reminders than on any reluctance to deal with AI. Anything complex can be passed to a buyer with the full conversation attached.

Yes. The agent is designed to persuade, not just push for the lowest price. It explains the reasoning behind each ask, responds to the supplier’s actual objections, and uses a tone suited to the category and relationship.

No. It handles the negotiations your team doesn’t have capacity for, so your people can focus on strategic deals and key supplier relationships. Your team sets the strategy and guardrails; the agent does the legwork.

For a narrow, well-defined negotiation, building can be reasonable if you have spare engineering capacity. The hard part is the negotiation itself: handling supplier pushback, staying within guardrails and improving over time takes a large volume of real negotiations to tune. That specialist capability is expensive to build and maintain in-house.

Specialist providers generally negotiate with suppliers at scale, within guardrails you set. Nibble differs in three ways:

• Focused only on negotiation. We build negotiation agents and the features they need, so Nibble plugs into your source-to-pay suite or orchestration layer rather than competing with it.
• Human-like negotiation. Suppliers negotiate in free text and the agent adapts, rather than offering a fixed set of options. Built on 2M+ real negotiations, with strategy developed with Professor Niro Sivanathan of London Business School.
• Beyond tail spend. Three agents on one platform cover tail-spend campaigns, single-supplier quotes and renewals, and competitive RFQs, so you save time on tactical spend as well as money.

Every offer stays inside your limits because the pricing logic sits outside the LLM, and anything complex goes to a buyer with the full conversation attached.

Usually not because the technology stops working, but because the pilot was never set up to be repeatable. A pilot run by one team on hand-picked data, with no owner for the next rollout, tends to stay a pilot. Pilots that scale have a focused first team, a clear internal owner, and agreement before launch on what success unlocks next.

Four questions cut through most demos. What happens if a negotiation goes outside your parameters: is there a hard guardrail, or does it rely on the model behaving? How many real negotiations has it run, rather than simulated ones? Can you see a full transcript and audit log for every negotiation? And how long, concretely, from demo to live pilot: weeks, or does it depend on a lengthy integration first?