UK BUYER’S GUIDE · 7 PRODUCTS
AI agent platforms
Can an agent complete a useful task, recognise a boundary and leave a reviewable record?
Compare agents that answer from business knowledge and perform controlled actions with visible results.
YOUR SHORTLIST STARTS HERE
Compare the options.
Highest documented score first. Fit adds up the 20 feature checks below, including AI, across documented plans. Entry prices may exclude scored capabilities. The upper end of a range represents unverified points.
7 of 7 products · Highest documented score first
| Product / good fit for | Pricing context | Documented fit | Next step |
|---|---|---|---|
Intercom and Fin Intercom and Fin Teams evaluating AI resolution alongside a shared human inbox. | US$0.99Fin outcome VAT basis not stated usage | 67–98/100 documented fit69% evidence coverage | Details ↗ |
respond.io respond.io Businesses organising a shared WhatsApp sales or support operation. | US$79.00month VAT basis not stated annual | 60–100/100 documented fit60% evidence coverage | Details ↗ |
Make Make Teams that need to inspect data movement and maintain complex connected processes visually. | US$0.00month within free allowance VAT not applicable free | 49–100/100 documented fit49% evidence coverage | Details ↗ |
n8n n8n Technical operations teams needing control over integration logic and deployment. | €20.00month VAT basis not stated annual | 42–100/100 documented fit42% evidence coverage | Details ↗ |
PolyAI PolyAI Organisations integrating conversational service into an existing contact centre. | Request pricingvoice minute and deployment scope VAT basis not stated quote | 41–100/100 documented fit41% evidence coverage | Details ↗ |
ElevenLabs Agents ElevenLabs Agents Teams prioritising conversational voice while retaining control of business integrations. | Request pricingconversation duration plus model usage VAT basis not stated usage | 25–100/100 documented fit25% evidence coverage | Details ↗ |
Zendesk Zendesk Service teams needing formal queues, reporting and room for operational controls. | £15.00agent/month VAT basis not stated annual | 19–100/100 documented fit19% evidence coverage | Details ↗ |
Prices retain their original currency, billing basis and VAT status. Check the evidence and plan requirements before buying.
| Capability & demo question | Intercom and Fin67–98/100 documented fit | respond.io60–100/100 documented fit | Make49–100/100 documented fit | n8n42–100/100 documented fit | PolyAI41–100/100 documented fit | ElevenLabs Agents25–100/100 documented fit | Zendesk19–100/100 documented fit |
|---|---|---|---|---|---|---|---|
Core workflowBusiness knowledge grounding7.0 points availableAsk in the demoAnswer a question using an approved document or help article. | ✓ Supported Fin uses approved policies and business information. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported AI Agents can answer configured FAQs. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported Attach business knowledge to ground the configured agent's responses. 7.0 / 7.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–7.0 / 7.0 points | ✓ Supported FAQs and sources ground responses. 7.0 / 7.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–7.0 / 7.0 points | ✓ Supported Agents use business knowledge to answer requests. 7.0 / 7.0 pointsEvidence ↗ |
Core workflowBusiness-system actions7.0 points availableAsk in the demoRetrieve or update a permitted record through an integration. | ✓ Supported Connectors, APIs and MCP perform external actions. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported Enabled skills update contact fields, lifecycle stages, tags and conversation assignment. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported Agents select and invoke connected workflows. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported Agents invoke workflow tools. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported Flows complete connected tasks. 7.0 / 7.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–7.0 / 7.0 points | ✓ Supported Advanced AI can complete service workflows. 7.0 / 7.0 pointsEvidence ↗ |
Core workflowMulti-turn task completion7.0 points availableAsk in the demoComplete a request that needs clarification over several turns. | ✓ Supported Procedures handle multi-step requests. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported Agents collect customer details through chat. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported A Conversation ID and history setting preserve context across turns. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported Attach memory for multi-turn chat; the documented default does not persist between sessions. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported Multi-step flows handle customer requests. 7.0 / 7.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–7.0 / 7.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–7.0 / 7.0 points |
Core workflowHuman escalation7.0 points availableAsk in the demoEscalate an unsupported request with its conversation context. | ✓ Supported Procedures support human handoff. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported Assign-to-agent or team skills transfer the conversation with context. 7.0 / 7.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–7.0 / 7.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–7.0 / 7.0 points | ✓ Supported Human handoff can preserve context. 7.0 / 7.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–7.0 / 7.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–7.0 / 7.0 points |
Core workflowConversation and action logs7.0 points availableAsk in the demoInspect the agent's messages and the actions it attempted. | ✓ Supported Fin provides operational monitoring and alerts. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported Test conversations show skill execution events and source snippets behind answers. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported The reasoning panel exposes decisions and tool use. 7.0 / 7.0 pointsEvidence ↗ | ✓ Supported Return intermediate tool steps and add tracing metadata for inspection. 7.0 / 7.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–7.0 / 7.0 points | ✓ Supported History and analytics expose conversations. 7.0 / 7.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–7.0 / 7.0 points |
Workflow depthKnowledge source management4.0 points availableAsk in the demoAdd, update and remove a source used by the agent. | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ✓ Supported Add or remove approved files and URLs; resynchronise website knowledge sources. 4.0 / 4.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points |
Workflow depthRestricted tool permissions4.0 points availableAsk in the demoLimit which systems and actions this agent can use. | ✓ Supported Restrict exposed response fields, eligible audiences and whether Fin may trigger a connector automatically. 4.0 / 4.0 pointsEvidence ↗ | ✓ Supported Select which AI skills an agent may execute in its configuration. 4.0 / 4.0 pointsEvidence ↗ | ✓ Supported Give the agent only explicitly attached modules, scenarios or other tools. 4.0 / 4.0 pointsEvidence ↗ | ✓ Supported Only configured tools are connected to the agent; selected tools require explicit human review. 4.0 / 4.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points |
Workflow depthApproval before consequential actions4.0 points availableAsk in the demoRequire a person to approve an action before it executes. | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ✓ Supported Manual approvals can be inserted. 4.0 / 4.0 pointsEvidence ↗ | ✓ Supported Gate selected tools behind approval; denial cancels the action and informs the agent. 4.0 / 4.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points |
Workflow depthFailed-tool handling4.0 points availableAsk in the demoSimulate a failed integration and show a useful recovery path. | ✓ Supported On authentication failure, dynamic tokens can refresh and retry once; failed refreshes are logged and repeated failures trigger a circuit breaker. 4.0 / 4.0 pointsEvidence ↗ | ? Not verified Human escalation is documented; a failed external tool's recovery path remains to be demonstrated. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points |
Workflow depthCustomer or user authentication4.0 points availableAsk in the demoProtect a private-record lookup from an unauthenticated visitor. | ✓ Supported Enable customer authentication before a connector accesses or displays private data. 4.0 / 4.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points |
Workflow depthOperational guardrails4.0 points availableAsk in the demoSet a restricted topic or action and test it. | ✓ Supported Required inputs and restricted response fields constrain connector use and exposed data. 4.0 / 4.0 pointsEvidence ↗ | ✓ Supported Configure allowed behaviours and escalation rules; test before deployment. 4.0 / 4.0 pointsEvidence ↗ | ✓ Supported Defined rules and stopping points constrain execution. 4.0 / 4.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points |
Workflow depthMultiple deployment channels4.0 points availableAsk in the demoUse the agent in two supported channels. | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ✓ Supported Agents work across WhatsApp and other messaging channels. 4.0 / 4.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ✓ Supported Voice and chat are supported. 4.0 / 4.0 pointsEvidence ↗ | ✓ Supported Agents support voice and text deployments. 4.0 / 4.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points |
Workflow depthVersions and environments4.0 points availableAsk in the demoTest a change separately before promoting it. | ✓ Supported Connector changes remain drafts until published; preview supports test-user validation. 4.0 / 4.0 pointsEvidence ↗ | ✓ Supported Simulate conversations and skill behaviour before publishing; real call transfer needs a live test. 4.0 / 4.0 pointsEvidence ↗ | ✓ Supported Test a configured agent before using it in a live scenario. 4.0 / 4.0 pointsEvidence ↗ | ✓ Supported Evaluate changed workflows on test inputs before deploying the revised workflow. 4.0 / 4.0 pointsEvidence ↗ | ✓ Supported Sandbox and Live separate changes. 4.0 / 4.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points |
Workflow depthRepeatable evaluation cases4.0 points availableAsk in the demoRun a saved set of cases after changing the agent. | ✓ Supported Saved tests and simulations check changes. 4.0 / 4.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ✓ Supported Run stored test datasets and compare quality across model or prompt changes. 4.0 / 4.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ✓ Supported Automated tests check agent behaviour. 4.0 / 4.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points |
Workflow depthTask-outcome analytics4.0 points availableAsk in the demoReport successful actions, escalations and failed requests separately. | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–4.0 / 4.0 points |
Workflow depthConversation-data export5.0 points availableAsk in the demoExport relevant logs for independent review. | ◐ Limited Conversation transcripts and AI metadata are retrievable, with a 500-part limit per conversation and feature-dependent AI fields. 2.5 / 5.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points |
Workflow depthUsage and cost visibility5.0 points availableAsk in the demoInspect usage and set or explain the spending limit. | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points |
AIModel choice or configuration5.0 points availableAsk in the demoShow how model behaviour or choice is configured for the task. | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ✓ Supported Configure instructions, role, behaviour and enabled skills for each agent. 5.0 / 5.0 pointsEvidence ↗ | ✓ Supported Choose the AI provider and available model in the agent module. 5.0 / 5.0 pointsEvidence ↗ | ✓ Supported Choose supported chat-model nodes and configure agent system instructions. 5.0 / 5.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ✓ Supported Choose a model or connect a custom LLM. 5.0 / 5.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points |
AINatural-language agent building5.0 points availableAsk in the demoDescribe an agent and inspect the generated configuration. | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ✓ Supported Wren creates an initial agent draft. 5.0 / 5.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points |
AIAI-assisted quality review5.0 points availableAsk in the demoIdentify weak conversations or score outcomes with configurable criteria. | ✓ Supported AI QA surfaces trends and improvement recommendations. 5.0 / 5.0 pointsEvidence ↗ | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ? Not verified Exact workflow not verified in the reviewed material; request a demonstration. 0.0–5.0 / 5.0 points | ✓ Supported Conversation analysis evaluates configured outcomes. 5.0 / 5.0 pointsEvidence ↗ | ✓ Supported QA add-on analyses conversations and knowledge gaps. 5.0 / 5.0 pointsEvidence ↗ |
“Not verified” means the recorded evidence does not settle this specific check. It does not mean the feature is absent. Read the scoring method.
Focus on up to three products →ASK BETTER QUESTIONS
Before you buy ai agent platforms software.
What to look for
- Define allowed actions and prohibited actions before loading documents.
- Maintain a small repeatable test set covering edge cases and handovers.
- Measure successful tasks and customer outcomes, not only chat volume.
What changes the cost
- Usage, tokens, outcomes or execution credits.
- Knowledge indexing, connectors and human seats.
- Integration engineering, evaluation and ongoing maintenance.
What to watch out for
- Giving an agent broad credentials to solve a narrow task.
- Assuming a fluent answer is grounded in your current policy.
- Counting a vendor's AI support assistant as a customer-deployable agent builder.
Common questions
What is the difference between a copilot and an agent?
For this comparison, an agent can perform a task using tools or structured workflows; drafting assistance alone is not treated as autonomous execution.
Why only 15 points labelled AI in an AI category?
The other 85 points assess the operational capabilities needed to run an agent reliably. The AI-labelled checks assess model, build and evaluation assistance specifically.
Does an agent need a human approval at every step?
No. Set approvals where the consequences warrant review, and test that restricted actions cannot bypass them.