Crescendo

What it is

Crescendo sells the half of AI customer service that most AI customer service vendors deliberately leave out: the people who handle what the agent cannot. It runs AI agents across voice, chat, email, SMS and messaging, and it staffs a live human operation underneath them — then bills both on the same per-resolution meter. Crescendo calls that combination augmented AI. Commercially it is a software vendor and a BPO sold as one contract, with one number on the invoice and one party accountable when a contact goes wrong.

The shape came from an acquisition, not a build. On 2 October 2024, Crescendo — then a roughly 20-person startup — acquired PartnerHero, adding over 200 customers and 3,000 CX professionals, with Airtable, Khan Academy, sweetgreen, Tom’s Shoes and Udemy among the accounts. The same announcement disclosed $50M raised over the prior year led by General Catalyst, with Celesta Capital and Alorica participating, at a reported $500M valuation.

Leadership changed on 14 July 2026. Co-founder Andy Lee — who founded Alorica and built it past 100,000 employees across 16 countries — became CEO, and founding CEO Matt Price moved to Strategic Advisor to the board. The services half of the business now has the services founder running it, which is the clearest signal available about which half sets the roadmap.

Why it shows up in CX stacks

• It answers the escalation question, not the deflection question. Every AI agent evaluation stalls at the same place: containment is 60-75%, and the remainder still needs a trained human on a channel, in a language, at an hour. Crescendo prices the AI-handled contact and the human-handled contact on one meter, so the fallback is contracted rather than hired. Buyers who cannot staff a 24/7 tier-1 queue are not choosing between Crescendo and Decagon; they are choosing between Crescendo and running a software pilot alongside an existing outsourcer contract.
• Multimodal inside a single conversation. At Enterprise Connect on 11 March 2026, Crescendo took both Best Innovation for Customer Experience and Overall Best of Show for letting a customer speak, type and share images or documents in one conversation without changing channel or losing context. For returns, damage claims and device troubleshooting, the photo is the ticket, and forcing that into an email thread is where handle time goes.
• The product consolidated in August 2026. On 25 August 2026 Crescendo launched the Crescendo Customer Experience Platform, folding CCaaS, ticketing, workforce management, quality assurance, Voice of the Customer and knowledge into one system with named AI agents running the work across them. Vendor-stated growth over the preceding year: AI-handled volume up 600%, AI customers up 200%.
• Time-to-live is contracted, not promised. The Optimization Agent, shipped 22 June 2026, claims a UAT-ready assistant in under an hour. The platform commitment is 30 days to go live, and the Total Outcomes Guarantee announced 23 October 2025 makes the entire implementation free if that date slips.
• Alorica resells it. Under a 23 June 2026 partnership, Alorica is licensed to resell and deploy the platform across its enterprise base, and Crescendo accounts get access to Alorica’s 100,000-plus staff across 16 countries. Crescendo states 500+ AI deployments worldwide.
• Europe is staffed, not just addressed. The UK and Europe launch landed 22 April 2026, on the back of passing $100M ARR in under two years, with London as the hub and a VP EMEA hired on 13 May 2026. Named European accounts: Dr. Martens, Funky Pigeon, Mention Me, RealVNC and Sweet Bee Organics.

Pricing reality

Nothing is published. The pricing page carries outcome claims and a demo button — no tiers, no rate card, no unit price.

The billing unit is a resolution rather than a seat or an hour, which is the part worth understanding before the first call. Third-party teardowns converge on two tiers: managed AI at roughly $2,900 per month plus about $1.25 per AI-solved resolution, and the AI-plus-human tier at the same monthly fee plus about $2.25 per resolution. G2’s pricing page lists $2.99 per resolution, described as including everything. All three figures are third-party estimates, none are published by Crescendo, and the monthly fee reportedly covers onboarding, QA, training, white-labelling and ongoing maintenance.

Normalise the unit before comparing anything. Intercom Fin bills $0.99 per resolution and Zendesk $1.50 per automated resolution, and both of those numbers buy software that your own team configures, tunes and staffs behind. Crescendo’s $2.25 buys the resolution and the person who takes it when the agent fails. Fin’s own comparison page runs the arithmetic at 100,000 monthly resolutions — $99,000 a month for Fin against $125,000 at Crescendo’s $1.25 and $299,000 at $2.99 — and the arithmetic is right, but it is competitor-authored and silent on the fact that the Crescendo line also retires an outsourcer invoice. The comparison that decides the deal is Crescendo’s per-resolution rate against your software cost plus your BPO’s per-contact rate plus the internal headcount managing both.

The Total Outcomes Guarantee makes three commitments: higher CSAT on AI-solved interactions than your incumbent or there is no charge, go live within 30 days or implementation is free, and payment only against positive outcomes. No numeric threshold is published for any of the three, which is the part to pin down in writing.

Best for

CX and support leaders at consumer, ecommerce and subscription brands running roughly 20,000 to 500,000 contacts a year, who need volume handled end to end and have no intention of building the internal team to run an agent platform, retrain it weekly and staff the queue behind it. It is the strongest pick when the alternative on the table is a software licence plus a separate BPO contract plus a CX ops hire, and when the buying committee wants one throat to choke for a containment number.

Skip it if support is a strategic surface you intend to own — a product-led company where the ticket stream feeds engineering, or a brand whose support voice is part of the product. Skip it if you already run a capable in-house team and only need the agent layer, if procurement cannot approve spend without a published rate card, or if contact volume sits below roughly 10,000 a year, where the monthly service fee dominates the bill.

Versus the alternatives

The volume incumbents are Zendesk and Intercom, and for most teams the cheapest first test is turning on the AI agent inside the helpdesk you already pay for — take that path when your ticket mix is simple, your knowledge base is current and you have someone to own it. Sierra is the fastest-growing entrant in the segment and, with Decagon, the right pick when you want to own and tune the agent yourself at enterprise volume. Cresta is the pick when the queue is voice-heavy and the goal is making human agents better rather than replacing the contact. Gladly wins where the customer rather than the ticket is the record. A traditional BPO plus any of the above is the honest baseline, and it is the one Crescendo has to beat: it is cheaper per contact in low-wage delivery geographies and it does not couple your software decision to your staffing decision. Crescendo’s argument against all of them is narrow and real — it is the only one where the AI vendor is contractually on the hook for the contact the AI fails to resolve.

Watch-outs

• You are outsourcing the operation, not buying a tool, and switching back is a rebuild. Knowledge, workflows, QA rubrics and agent training live with the vendor. Two years in, the institutional knowledge your team would have accumulated from the ticket stream sits in Crescendo’s system instead, and an exit means hiring, tooling and retraining at once. Guard: contract for export of the knowledge base, conversation transcripts and agent configuration in a usable format at any time, not at termination, and run a quarterly internal review of the top 20 contact drivers so your own team keeps reading the stream.
• Every headline outcome number is vendor-published and each defines its own metric. 75% of enquiries resolved at Stewart Golf, 85% instant resolutions and a 34% CSAT lift at IDEO U, up to 99% accuracy at Rachio, 600% growth in AI-handled volume — all Crescendo material, none independently audited. Guard: make the acceptance test your own definition of resolution measured on your own traffic over a two-week shadow period, and write that definition into the contract rather than into the statement of work.
• The pricing is unpublished and the meter is one you do not control. Crescendo decides what counts as a resolution and its own agents generate the volume being billed. Guard: define a resolution in the contract — no reopen within seven days, no immediate re-contact on the same issue, no credit for a transfer — cap billable resolutions per contact, and require a monthly line-item report you can reconcile against your own CRM.
• The company has changed shape three times in two years, and the platform is three weeks old. A 20-person startup became a 3,000-person services business in October 2024, changed CEO in July 2026, and shipped the consolidated platform on 25 August 2026. The CCaaS, workforce management and QA modules you will be demoed have almost no production track record at any scale. Guard: buy the agent and human-resolution layers on contract and take the newer modules as a no-cost pilot addendum with a named exit, and ask for reference customers at your volume who have been live on the consolidated platform for at least two quarters.
• A human operation handling your customers is a data-processing and employment-law surface, not just a vendor line. Contacts are handled by staff across six continents, which pulls GDPR transfer mechanisms, sub-processor disclosure and works council consultation in DACH and France into a decision that looks like a software purchase. Guard: get the full sub-processor list and delivery geographies in writing before the pilot, complete a DPIA covering both the AI processing and the human handling, and confirm which geography handles EU customer contacts rather than assuming it follows the contracting entity.