Capacity

What it is

Capacity is a St. Louis-based CX automation platform, founded in 2017 by David Karandish, that sells AI agents across chat, voice, SMS, email and social, plus real-time agent assist, automated QA, conversation intelligence, a helpdesk, live chat, a knowledge base, outbound campaigns, scheduling and payments. All of it sits on one shared knowledge layer that every agent, bot and human reads from. Most of that catalogue was bought rather than built: Capacity has made more than a dozen acquisitions since 2023, including Textel (business texting), LumenVox (speech recognition), SmartAction (voice virtual agents), CereProc (text-to-speech), Creovai (agent assist and QA), Starmind (expertise mapping, March 2026) and Chattigo (Latin American omnichannel messaging, July 2026).

The business is real and growing. CMSWire reports ARR passed $100M on June 18, 2026, more than 20,000 customers, and a Series E of over $54M announced September 2, 2026, taking total capital past $159M. The closest comparison is not Intercom or eesel but a contact-centre suite: Capacity competes for the same budget as NICE, Genesys and Five9 add-ons, aimed at mid-market and enterprise contact centres.

Why it shows up in Customer Success stacks

• One vendor instead of four. A team that would otherwise buy a helpdesk, a voice bot, an agent-assist tool and a QA product can buy them on one contract with one knowledge base. That is the case Capacity makes, and it is the only reason to shortlist it over a point tool.
• Voice is native, not a partner. The LumenVox, SmartAction and CereProc acquisitions mean speech recognition, voice agents and text-to-speech are Capacity’s own stack. Most chat-first AI agents (Fin, eesel, Forethought) treat voice as an add-on or skip it.
• It sits beside an existing CCaaS. Capacity integrates with Five9, Genesys Cloud, Salesforce, ServiceNow and HubSpot, so a contact centre can add AI agents and assist without ripping out its telephony.

Pricing reality

Capacity publishes structure, not prices. The model has three parts:

• Platform fee: an annual subscription in three tiers. Core includes the knowledge base, AI agent builder, workflows, LLM hub and basic analytics. Pro adds the helpdesk, live chat and inbox. Enterprise adds campaigns, CRM and an optional HIPAA package.
• Usage: AI agents bill per chat response, per voice minute, per email response and per SMS message. Agent assist and conversation intelligence bill per hour of agent time. LLM token costs are included.
• Add-ons and services: integrations, custom implementation and managed services are scoped in the proposal.

The only public number is Capterra’s starting price of $25,000/year for the entry tier, and that is before a single AI response is billed. G2 buyers report an average discount of about 12% off list and rate perceived cost at the top of G2’s five-point scale. Treat $25K as the floor for a single-channel pilot, and expect a multi-channel deployment with assist and QA to run several times that once usage is added.

The detail that matters when modelling cost: chat is billed per response, not per conversation. A chat that takes four bot replies is four billable events. Fin bills $0.99 per resolved conversation and eesel bills per ticket handled, so ask Capacity to quote on a blended per-conversation basis using your own average replies per chat, or you cannot compare the bids.

Best for

Contact-centre and CX operations leaders at mid-market and enterprise companies running 100+ agents across voice and digital channels, who want to replace several point tools (voice bot, agent assist, QA, texting) with one vendor and one knowledge base, and who have the staff to run a three-month implementation.

Not for SaaS support teams under about 30 agents that only need chat and email deflection, or any team that wants published prices and a self-serve trial. A chat-first agent on your current helpdesk will be cheaper and live in days.

Watch-outs

• Implementation is long and partner-led. G2 buyers report about 3 months to implement and 14 months to ROI; Capacity’s own FAQ puts voice at 6 to 8 weeks to production. On a 12-month term, renewal comes before payback. Guard: negotiate a 24-month term with a 90-day exit if go-live slips past an agreed date, or a first-year ramp where usage is discounted until launch.
• Acquisition-built products can feel stitched together. Thirteen-plus acquired codebases do not become one product overnight; G2’s auto-summarised cons cluster on complex setup. Guard: in the demo, require one end-to-end scenario (voice call → AI agent → handoff with assist → auto-QA score) running in a single console on your data, not a module-by-module tour.
• Helpdesk integrations vary in depth. Third-party reviews of Capacity’s directory list Zendesk and Freshdesk as API-only rather than indexed knowledge with an embedded UI. Guard: if your agents live in Zendesk or Freshdesk, test whether Capacity can read your macros and help-centre articles directly before signing; otherwise budget for its own helpdesk on the Pro tier.
• Per-response metering makes bills hard to predict. Guard: cap annual usage in the contract with a fixed overage rate, and get a per-channel usage report monthly so verbose bot flows show up before the invoice.

Alternatives

Zendesk and Salesforce Agentforce are the market-share leaders: pick either when your agents already work in that system and you want AI agents inside it, accepting per-resolution pricing around $1.50 to $2.00 on Zendesk. Intercom Fin is the pick for chat-and-email-first SaaS support at $0.99 per resolution with no platform floor. Sierra and Decagon are the fastest-growing entrants for enterprise AI agents with complex backend actions; pick them when AI agent quality is the whole decision and you do not need Capacity’s helpdesk, texting and QA in the same contract. Ada and Forethought sit between: AI-agent specialists that plug into your existing helpdesk. Pick Capacity when you want to consolidate voice, digital, assist and QA onto one knowledge layer and can absorb a $25K-plus floor and a quarter-long rollout.