How bouts are scored

Where the facts come from

Prices come from each vendor's public pricing pages, in USD list prices with annual billing where offered. Features, AI and security come from the vendor's own documentation, trust center and data processing terms. User ratings come from G2 and Capterra, recorded with the number of reviews and the date we saw them. Vendor claims about their own results never decide a round. We have not run these products on our own support queue.

Round 1: The bill

Who costs less once the AI is switched on? The lower monthly bill wins if it is at least 10% lower; closer than that, the round is even. A bill that leaves out an AI price the vendor doesn't publish can't beat a complete bill.

Round 2: Price tags

Whose full price can you work out without a sales call? The more open price list wins; the same tier is even.

1. No prices at all
2. An entry price and little else
3. Seat prices, with the AI priced by quote or not at all
4. AI prices in public, but with a minimum, annual-only terms or vague limits
5. Every price in public, monthly billing and no minimum

Tiers from lowest (1) to highest (5). The higher tier wins the round.

Round 3: AI

Whose AI does more for customers and agents? The more capable AI, going by each vendor's documentation, wins; the same tier is even.

1. No AI
2. AI that helps agents write, but nothing customer-facing
3. An AI agent that answers from the help center
4. An AI agent that also takes actions in other systems
5. An action-taking AI agent on chat, email and phone, with test runs and reporting

Tiers from lowest (1) to highest (5). The higher tier wins the round.

Round 4: Desk basics

Who covers more of a support team's daily work? The fuller set of channels and workflow tools wins; the same tier is even.

1. Email only
2. Email plus one more channel
3. Email, chat, a help center and SLAs
4. All that plus phone and messaging apps
5. Every channel natively, skills-based routing and 300+ integrations

Tiers from lowest (1) to highest (5). The higher tier wins the round.

Round 5: Security

Whose data handling is easier to defend to an auditor? The stronger set of audit reports and data promises wins; the same tier is even.

1. Nothing published
2. Security claims without audit reports
3. One audit report (SOC 2 Type II or ISO 27001)
4. SOC 2 Type II and ISO 27001, plus EU hosting, a HIPAA BAA or a no-training promise
5. SOC 2 Type II, ISO 27001, EU hosting, a HIPAA BAA and a no-training promise

Tiers from lowest (1) to highest (5). The higher tier wins the round.

Round 6: User verdict

Who do users rate higher? The higher G2 rating wins if the gap is 0.1 or more and both have at least 100 reviews there. Otherwise Capterra decides on the same terms. If neither settles it, the round is even.

The bill, in detail

We price the cheapest plan that includes email and live chat, runs the vendor's own AI agent and can add its agent assist, and keep that plan at every team size. Each agent handles 250 conversations a month and the AI agent resolves 20% of them; agent assist is bought for every agent. Taxes, phone minutes and one-off onboarding fees are left out. Weight classes: small team (3 agents, 750 conversations a month); mid-size team (10 agents, 2,500 conversations a month); large team (25 agents, 6,250 conversations a month).

Results

A win by four or more rounds is a knockout, by two or three a clear win, and by one a close call. Equal rounds won is a draw. The bout page headline uses the mid-size team; the other weight classes are shown beside it.

Rankings order help desks by wins, with a draw counting as half a win; ties go to the better difference between rounds won and lost.

Independence

No vendor pays for a result, a bout or a position, and we use no affiliate links. Vendors can send corrections with a public source.