Sierra pricing

About

Sierra is an enterprise conversational-AI agent platform founded in 2023 by Bret Taylor (former co-CEO of Salesforce and chair of the OpenAI board) and Clay Bavor (former Google VP). It bills itself as “the leading conversational AI platform for business” and lets enterprises deploy a single agent across chat, SMS, WhatsApp, email, voice and ChatGPT. Named customers include Rocket Mortgage, SiriusXM, Brex, Redfin, ADT and SoFi, spanning financial services, telecom, healthcare, media, travel and retail. Sierra carries SOC 2, ISO 27001, HIPAA, GDPR and FedRAMP credentials.

The product line as of July 2026 has four named surfaces: Ghostwriter (the “agent-building agent” — Build and Optimize), Agent Studio, Insights (Explorer, Monitors, Experiments, Observability) and Horizon, which launched on 2026-07-16 and replaced the former Agent Data Platform. Horizon extends Sierra from single conversations to long-horizon outcomes — agents that “plan and execute the work over weeks or months” toward a goal such as a mortgage, a claim or a renewal. Sierra’s homepage headline moved in step, from “Better customer experiences” to “Better outcomes. Built on Sierra.”

Sierra is among the fastest-scaling enterprise-AI companies on record: it raised 175M USD at a 4.5B USD valuation in October 2024, 350M USD at a 10B USD valuation in September 2025, and roughly 950M USD in May 2026. It crossed 100M USD ARR in late 2025 — about seven quarters after its February 2024 launch — and its own year-two review put it at “over $150M in ARR” entering year three in February 2026, after “our first-ever $50M quarter.”

For the most current information, visit Sierra.

Pricing summary : How Sierra’s pricing model works

Sierra is the flagship of outcome-based pricing — and it is sales-only. There is no public rate card: sierra.ai/pricing returns “Page not found.” The commercial model is communicated as a philosophy on the homepage instead — “Pay for a job well done” / “Ensure you only pay for the value Sierra delivers with outcome-based pricing.” In practice, the dominant meter is per resolution: Sierra is paid when its AI agent autonomously resolves a customer issue, not per seat and not merely per conversation.

The billing dimensions Sierra’s own surfaces name or imply, plus one that is only ever reported:

• Resolutions (outcome meter) — the headline unit; charged when the agent completes the job. No dollar figure is published.
• Interaction volume — volume-based pricing for routine, simple traffic, blended with the outcome meter for complex inquiries. This blend is reported by market research, not stated on any Sierra surface; Sierra’s own pages name only the outcome meter.
• Long-horizon outcomes (new, July 2026) — Horizon prices a goal that unfolds over days, weeks or months (“a mortgage, a claim, a renewal”) rather than a single conversation, and states the frame explicitly: “It’s Sierra’s outcomes-based model at scale: you pay for results, not tokens.”
• Channel mix — one agent spanning chat, SMS, WhatsApp, email, voice and ChatGPT; voice carries its own economics.
• Annual enterprise commitment — invoiced contracts; no self-serve checkout and no public plan picker on any Sierra surface. Contract minimums are not published, so any floor is a negotiated term rather than a stated one.

What makes this different: Sierra is the most visible champion of the thesis that AI agents should be paid like outsourced labor — for results, not access. The Horizon page sharpens that into an explicit rejection of the token meter: you pay for results, not tokens — and it stretches per-resolution pricing past the single ticket into outcomes measured in weeks.

Pricing by product

Sierra platform (commercial models)

Sierra publishes no rate card at all — sierra.ai/pricing returns “Page not found. Sorry, but the page you’re looking for might have been moved, renamed, or no longer exists.” Every price is quoted.

Horizon (long-horizon outcome agents, launched 2026-07-16)

Insights (agent analytics — the measurement layer under the meter)

Sales motions across products: Sierra is entirely sales-led — outcome-based and quoted per enterprise. No per-resolution or per-outcome rate is published on any Sierra surface; the rows above describe structure, not list prices.

Hidden costs : What Sierra users actually pay

With no published rates, Sierra’s cost drivers are contractual and definitional:

• Resolution definition — under per-resolution pricing, what counts as a resolved “outcome” (vs. an escalation or partial answer) is the central negotiation; it directly determines the bill.
• Blend ratio — the split between volume-priced routine interactions and outcome-priced complex ones shapes effective cost per contact.
• Channel mix — voice deployments generally carry different economics than chat/messaging.
• Annual commitment / minimums — enterprise contracts are annual with floors.
• Implementation — integrating Sierra’s agents with backend systems and knowledge is part of the engagement.

Want to estimate your own Sierra bill? Use the Sierra pricing calculator to model per-resolution outcome scenarios.

Pricing evolution : Sierra pricing history and changes

Cadence

Tracked range: 2024 Q1–2026 Q3. Quarters not listed above were verified stable (0 price changes, 0 SKU additions). Sierra has never published a per-resolution or per-outcome rate, so every entry above is a change in structure or posture, never in price.

Notable changes

• 2024-02 — Public launch foregrounding outcome-based pricing; the “pay for a job well done” claim shipped in place of, not alongside, a rate card.
• 2024-10 — 175M USD raise at a 4.5B USD valuation. Pricing posture unchanged; the outcome meter remained the resolved support conversation.
• 2025-09-04 — 350M USD led by Greenoaks at a 10B USD valuation, roughly a 100× multiple carried substantially by the outcome-pricing story rather than by disclosed unit economics.
• 2025-11 — Crossed 100M USD ARR in under two years — evidence that enterprises will sign eight-figure contracts against a meter whose rate is never published.
• 2026-05-04 — ~950M USD raise. No change to the model or to the absence of a pricing page.
• 2026-07-16 — Horizon launches, replacing the Agent Data Platform line. The billable outcome moves from a resolved conversation to a business goal that spans days, weeks or months, and the product page states the commercial frame outright: “It’s Sierra’s outcomes-based model at scale: you pay for results, not tokens.” No price changed, because Sierra publishes none.

The Horizon unit shift in detail

The 2026-07-16 launch is a repricing event with no price in it — which is exactly why it matters. Everything Sierra changed sits on the left side of the invoice: what a billable unit is.

• Before, the meter’s atom was a resolved conversation. A ticket arrived, the agent handled it end to end, and that resolution was the thing Sierra could invoice. Agent Data Platform — memory, customer data, recommendations, proactive engagement — was supporting infrastructure underneath that atom.
• After, Horizon’s Playbooks make the atom a business goal: “Give agents an outcome — a mortgage, a claim, a renewal — and they will plan and execute the work over weeks or months.” A single billable unit can now span dozens of interactions across phone, SMS and back-office systems, as the healthcare walkthrough on the Horizon page shows over a Day 1 / Day 2 arc.
• The economic consequence is that unit price and contract value both rise, and both become harder to audit. A resolved ticket is cheap, frequent, and easy to count. A booked appointment or a closed mortgage is expensive, rare, and arrives after an attribution window measured in weeks — during which competing touchpoints, human agents and the customer’s own behaviour all contribute.
• The controls Sierra shipped alongside it are volume-governing, not reporting. Built-in suppression avoids customers “already in progress, opted out, or no longer relevant”; Auditing and analysis exists to show “what triggered every engagement, what the agent decided, and why anyone was suppressed.” Under outbound, long-running agents, a suppression rule is a spend control — and Sierra built the audit trail for it before any buyer could ask.
• “You pay for results, not tokens” is the load-bearing sentence. It is a competitive claim against token-metered agent platforms, and it is also a margin commitment: Sierra is absorbing inference cost for work that now runs for weeks, on model prices it does not set.

What’s unique : Sierra’s distinctive pricing mechanics

1. Outcome pricing as identity. Sierra didn’t just adopt per-resolution pricing — it built its brand around it, with a founder evangelizing the model industry-wide.

2. Blended meters. Volume pricing for the easy stuff, outcome pricing for the hard stuff, lets Sierra align price with delivered value across a spectrum of complexity.

3. Labor-replacement framing. “Pay for a job well done” positions the AI agent as outsourced labor priced on results — a deliberate contrast to per-seat SaaS.

4. The unit of value moved up, not the price. Most vendors respond to rising AI capability by raising rates or adding a token line. On 2026-07-16 Sierra did neither: Horizon left the rate card blank (there isn’t one) and instead redefined the atom being sold, from a resolved conversation to a mortgage, a claim or a renewal that takes weeks. The product page names the alternative meter in order to reject it — “you pay for results, not tokens” — which puts model-cost volatility on Sierra’s side of the contract and bets on inference prices continuing to fall. Increasing contract value by enlarging the unit rather than by charging more per unit is the rarer of the two moves, and it is why a launch with no numbers in it belongs in a pricing blueprint. Compare with per-resolution pricing peers that quietly pass token cost through.

5. Suppression is a spend control, and Sierra shipped it first. Long-horizon outbound agents can manufacture their own billable volume, so the honest test of an outcome meter is whether the vendor gives buyers a brake. Horizon’s Built-in suppression skips customers “already in progress, opted out, or no longer relevant,” and Auditing and analysis explains “why anyone was suppressed.” Sierra built the negative-billing control into the launch rather than adding it after the first disputed invoice — see choosing the right usage metric for why that ordering matters.

Strengths & weaknesses

Billing UX : Sierra billing controls and transparency

Sierra ships no public billing console, but it names the controls that govern what becomes billable:

• Insights › Observability — “Understand every agent action — from tool calls, knowledge lookups, latency and more.” This is the audit trail buyers reconcile outcome charges against.
• Insights › Explorer — “ChatGPT-style Deep Research for conversations,” used to analyse agent performance and, in practice, resolution rate.
• Insights › Monitors — proactively “identify conversations needing extra attention,” i.e. the ones at risk of not resolving.
• Insights › Experiments — multivariate tests “to optimize conversation design and agent performance,” the lever that moves effective cost per contact.
• Horizon › Auditing and analysis — “See exactly what triggered every engagement, what the agent decided, and why anyone was suppressed.” The clearest outcome-attribution control Sierra publishes.
• Horizon › Goals and guardrails — “Define what the agent can do on its own, and where human sign-off is required,” which bounds autonomous, billable action.
• Horizon › Built-in suppression — avoids reaching customers “already in progress, opted out, or no longer relevant,” suppressing touches that would otherwise generate interactions.
• Horizon › Consent-aware delivery — “Enforce consent and opt-out automatically, per customer and per channel, on every single touch.”
• Payment options — invoiced annual enterprise billing; there is no self-serve checkout, no public plan picker, and no pricing page (sierra.ai/pricing 404s).

Strategic wins : Why Sierra’s pricing decisions worked

1. Owning the outcome-pricing narrative

By making outcome-based pricing its public identity, Sierra became the reference point for the category — see the outcome-based pricing revolution and how AI companies are shifting from per-user licenses.

2. Aligning price with delivered value

Charging per resolution lowers buyer risk — you pay when the AI works — which shortens enterprise sales cycles for an unproven technology. See choosing the right usage metric.

3. Blending meters for the full complexity range

Volume pricing for routine traffic plus outcome pricing for hard cases captures value without overcharging for trivial deflections. See the introduction to usage-based pricing.

4. Growing the meter upward instead of adding a token line

The obvious way for an AI vendor to monetise more capability is to bolt on consumption billing. Sierra’s 2026-07-16 Horizon launch took the opposite route: it enlarged the unit — a mortgage, a claim, a renewal instead of a ticket — while explicitly refusing the token meter. That keeps the invoice denominated in something the buyer’s own P&L already recognises, which is the whole reason outcome pricing shortens enterprise sales cycles in the first place. See how AI companies are shifting from per-user licenses and the case for value metrics over consumption metrics.

5. Shipping the volume brake with the product

Outbound, multi-week agents can generate their own billable activity, and buyers know it. By launching Horizon with Built-in suppression, consent-aware delivery and an audit trail that explains “why anyone was suppressed,” Sierra pre-empted the objection that outcome pricing rewards a vendor for reaching out more. Publishing the control before the rate is an unusual sequencing choice, and it is the credible half of the “results, not tokens” claim.

Areas to improve : Gaps in Sierra’s pricing approach

1. Opacity

No published rates means every prospect enters a sales cycle and finance teams cannot self-estimate — a friction point covered in bill shock and cost unpredictability.

2. Outcome definition risk, now stretched over weeks

Disputes over what counts as a resolved outcome already sat at the centre of every Sierra negotiation. Horizon (2026-07-16) widens the exposure: when the billable unit is a booked appointment or a closed mortgage that took a fortnight, a human agent, a branch visit and the customer’s own initiative all contributed to it. The concrete fix is a published attribution standard — an attribution window, a stated rule for partial credit when a human closes the loop, and a documented definition of “outcome” distinct from “resolution.” Sierra already ships the audit trail (Horizon › Auditing and analysis); what is missing is the contractual definition the trail is measured against.

3. Forecastability across multi-month outcomes

Per-resolution billing already fluctuated with volume and resolution rates. A meter whose unit completes over weeks or months adds a lag problem on top: spend in month one produces invoices in month three, which is awkward for both budget owners and revenue recognition. The fix is a forecasting artefact, not a discount — an in-product view of in-flight Horizon outcomes and their expected completion, so finance can accrue against work already underway rather than discovering it at settlement. See bill shock and cost unpredictability and revenue recognition for usage-based models.

4. No published unit taxonomy

After the Horizon launch Sierra names two different things an agent can complete — a resolved conversation and a long-horizon business outcome — but no public surface says whether they are one meter or two. A buyer cannot tell from sierra.ai whether a mortgage counts as one outcome or as the dozens of interactions it contains. The fix is cheap and does not require publishing a rate: a single page naming the billable units and their relationship. Vendors with sales-only pricing can still be structurally transparent, and the gap here is definitional, not commercial.

Key takeaways

1. Grow the unit before you grow the rate. Sierra’s July 2026 Horizon launch added contract value by redefining the billable outcome from a ticket to a mortgage, not by raising a price. When your product gets more capable, ask what larger job it can now be paid for before you ask what to charge per job.
2. Naming the meter you refuse is a positioning move. “You pay for results, not tokens” only lands because token billing is the default everyone expects. Stating what you are not charging for is often clearer to a buyer than describing what you are.
3. Blend meters to match complexity. Volume pricing for routine traffic and outcome pricing for hard cases lets one contract cover the full difficulty range without overcharging for trivial deflections.
4. Ship the spend brake with the launch. Suppression, consent enforcement and an audit trail arrived with Horizon rather than after the first disputed invoice — which is what makes an outcome meter survivable for the buyer who has to defend it internally.
5. A missing rate card is not the same as a missing structure. Sierra sells eight-figure contracts with no published price, but it publishes the shape of the model relentlessly. Buyers tolerate opacity about the number far better than opacity about the mechanic.

UBP implications

1. Outcome units are not fixed — they ratchet upward. Sierra spent two years establishing the resolved conversation as a billable atom, then in July 2026 replaced it with an outcome that takes weeks. Any UBP model built on an “outcome” should expect that unit to be renegotiated upward as the agent’s autonomy grows, and should price the first unit knowing a bigger one is coming.
2. Attribution windows are the next contract battleground. Per-resolution pricing settles within a session, so attribution was trivial. Once an outcome unfolds over weeks alongside humans and other channels, the industry needs an attribution standard — window length, partial credit, exclusion rules — the way performance marketing eventually needed one. Definition is still the contract; the definition just got harder.
3. “Results, not tokens” is a margin bet as much as a pricing model. A vendor who charges for outcomes and refuses to pass through inference cost is underwriting model prices it does not control. That is defensible while inference deflates, and it is the clearest reason outcome pricing has clustered among the best-capitalised AI vendors rather than spreading evenly across the category.

Sources

• Sierra official website (accessed 2026-07-22)
• Sierra — Horizon product page (accessed 2026-07-22)
• Sierra pricing page (returns “Page not found”) — re-confirmed HTTP 404 (accessed 2026-07-23)
• Sierra — Product overview — “Only pay for the value delivered”; no rate card on any product surface (accessed 2026-07-23)
• Sierra — Agent Studio — no pricing or dollar figures (accessed 2026-07-23)
• Sierra blog — The next Horizon in agents — “with Horizon, you don’t pay for tokens, you pay for business outcomes delivered” (accessed 2026-07-23)
• Sierra blog — Year two in review — “over $150M in ARR” (accessed 2026-07-23)
• Sierra — AI agents, outcome-based pricing, and the OpenAI board (podcast) (accessed 2026-06-11)
• Sierra blog (accessed 2026-07-23)

Bottom line

Sierra is the highest-profile outcome-based AI agent vendor: you pay per resolution — “a job well done” — blended with volume pricing for routine interactions, and, since Horizon shipped on 2026-07-16, for business goals that take weeks rather than minutes. It still publishes no rates and sells only through enterprise sales, so the buyer’s real negotiation is definitional: what counts as an outcome, and over what window. Sierra answered the second question with product (suppression, consent, an audit trail) before answering the first with contract language — which is the most interesting thing about this launch, and the thing to watch next. Browse the pricing blueprint for fully-researched company profiles.

Want to compare Sierra against other customer-service AI companies? Browse the pricing blueprint.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Horizon launches — outcome pricing extended to long-horizon agents

Jul 2026

Sierra introduced Horizon, replacing the Agent Data Platform product line. Horizon agents plan and execute an outcome (a mortgage, a claim, a renewal) across days, weeks or months, and the product page states the commercial frame directly: 'It's Sierra's outcomes-based model at scale: you pay for results, not tokens.' Pricing stayed sales-only with no published rate.

~950M USD raise

May 2026

Sierra raised roughly 950M USD as the enterprise-AI capital race intensified; pricing remained outcome-based and quote-only.

10B USD valuation

Sep 2025

Sierra raised 350M USD led by Greenoaks at a 10B USD valuation, more than doubling its 4.5B USD October 2024 mark; the outcome-based model remained sales-only.

4.5B USD valuation on an unpublished meter

Oct 2024

Sierra raised 175M USD at a 4.5B USD valuation roughly eight months after launch. The billable unit remained a single resolved conversation and no rate card appeared, establishing the pattern that Sierra publishes the shape of its model but never the number.

Sierra launches with outcome-based pricing

Feb 2024

Sierra publicly launched its enterprise conversational-AI agent platform, foregrounding outcome-based pricing — 'pay for a job well done' — as a category-defining alternative to per-seat software.