Explore AI by task, platform, industry and company.
This directory is the practical index for Top AI Chatbots. We organise AI products by what they do, where they operate, who they serve and the companies behind them.
New vertical · 1 November 2026 · AI IT helpdesk
Best AI IT helpdesk agents: Serval vs Moveworks vs ServiceNow vs Freshservice
“Have you tried restarting?” is not a resolved IT ticket. The question is whether an AI agent can fix the issue, follow access rules, update the record and hand over when stuck. On 29 October, Serval announced a $150m Series C at a $3bn valuation and launched Simulations to test its helpdesk agent against realistic tickets before changes go live. Reuters independently covered the funding on 29 October. Funding is not a product-quality score.
Start with Serval if you want to investigate an AI-native ITSM replacement; Moveworks if you want an employee-facing assistant over existing tools; ServiceNow if your enterprise already runs its ITSM platform; and Freshservice for a more conventional ITSM suite with published base pricing. These are fit-based recommendations from vendor evidence, not hands-on tests.
Pick your fighter: what is the job?
Choose a platform below to jump to its profile.
⚡ Serval ↓Build, test and run IT workflows with controlled approvals.
💬 Moveworks ↓Employee self-service over existing ticket and identity systems.
🛠 ServiceNow ↓Native AI specialists within an established ITSM platform.
📋 Freshservice ↓ITSM suite with Freddy AI options and plan caveats.
| Product | What it is | Ask for proof of |
|---|---|---|
| Serval | AI-native service management and workflow builder | Simulation, approval and real ticket closure |
| Moveworks | Assistant layered over existing ITSM | Identity actions, ticket updates and handoff |
| ServiceNow | ITSM platform with L1 AI specialist | Contract eligibility, audit logs and CMDB context |
| Freshservice | ITSM suite with Freddy AI | AI plan entitlement, sessions and total price |
Serval — best to investigate for AI-native workflows
Great if you want to rethink the ITSM system rather than add a chat window. Serval’s Catalyst builds workflows from natural language or ticket history, with review and approvals. Its 29 October Simulations use mock users and systems to test outcomes without making real access changes by default. Skip it if you only need a low-effort FAQ bot. Product · Simulations.
Moveworks — best to investigate as an employee front door
Great if your employees use Slack or Teams and your ticketing system is already in place. Moveworks documents password resets, troubleshooting, ticket updates and approval trails across systems. Bad fit if you expect to remove the underlying ITSM entirely. ServiceNow acquired Moveworks; compare standalone and bundled offers carefully. Official product.
ServiceNow — best first evaluation for existing customers
Great if your incidents, assets and changes already live in ServiceNow. Its L1 Service Desk AI Specialist is designed for support requests such as password resets and software access. Bad fit if your budget cannot support the wider deployment. Confirm feature availability and licensing in writing; ServiceNow’s performance figures are vendor-reported. Official announcement.
Freshservice — best to compare when predictable ITSM matters
Great if you need established ticketing, service catalogues and SLAs. Its US pricing page lists Starter at $19, Growth $49 and Pro $99 per agent/month billed annually; Enterprise is quote-based. Freddy AI Agent (Classic) is listed under Enterprise and Copilot may cost extra. Bad fit if you assume the $19 headline buys autonomous resolution. Request the complete AI quote.
The six-ticket test before you buy
- Verify a real password reset after checking identity.
- Request privileged access as a contractor; require sponsor approval first.
- Ask about a device belonging to another employee; prevent mistaken changes.
- Break the identity integration in a sandbox; confirm safe failure and escalation.
- Change a workflow; run regression tests before publishing it.
- Request a human; inspect the ticket history, handoff and audit trail.
Measure verified resolution rate, permission violations, repeat tickets, mean time to resolution and cost per completed request. A friendly answer is not a closed ticket. Use our AI agent evaluation checklist.
Quick answers
Is Serval the same as Moveworks? No: Serval positions itself as an AI-native service platform, while Moveworks is an assistant that can sit over existing ITSM. Does ServiceNow own Moveworks? Yes; compare its standalone and integrated offerings. Which is cheapest? Public base-seat prices do not establish the cost of autonomous AI. Get comparable implementation, licence and usage quotes. Can agents grant access safely? Only with independent identity checks, approvals, least privilege, logs and tested rollback.
Independent AI discovery, not recycled funding announcements.
Related: Agent evaluation · Personal agents · Buying guides · Our methodology.
Research checked 1 November 2026. Independent editorial analysis of vendor documentation and Reuters reporting; no hands-on deployments, affiliate commissions or paid placements claimed.
New comparison · 31 October 2026 · AI spreadsheets
Best AI spreadsheet assistants: Gemini Sheets vs Copilot Excel vs ChatGPT vs Rows
The real question isn’t “can AI write a formula?” It’s can it edit your actual workbook, explain what changed and leave the numbers correct? This week made the choice more interesting: Google’s 29 October Workspace update adds web citations and up to 3,000 cells per request for Fill with Gemini. Meanwhile Microsoft’s experimental =COPILOT() worksheet function stopped working on 14 September; Copilot in Excel’s side pane remains available.
Already work in Excel? Start with Microsoft Copilot in Excel. Live in Google Workspace? Try Gemini in Sheets. Need to interrogate an uploaded CSV or XLSX and produce charts? ChatGPT is the practical starting point. Need a new spreadsheet workspace with connected sources and built-in AI? Investigate Rows. These are capability-based recommendations, not results of our own hands-on benchmark.
Pick your fighter: where does the data live?
✨ Gemini in Sheets →Best first look for Sheets teams: cell filling, source citations and connected Workspace tasks.
🧠 ChatGPT →Upload CSV/XLSX files, ask questions, inspect code-backed analysis and create charts.
⚡ Rows ↗Build connected tables, enrich data and make reports with its AI Analyst.
What each tool actually changes
| Tool | Useful strength | Watch out for |
|---|---|---|
| Copilot in Excel | Edit workbooks using Excel-native tools | Licensing and supported file requirements; old =COPILOT() formulas now return #NAME? |
| Gemini in Sheets | Fill cells with cited web research; Sheets canvas | Rollout, Workspace entitlement and whether generated facts match the source |
| ChatGPT | Analyse uploads, calculate, chart and explain findings | Uploaded snapshots may not stay synced with a changing workbook; inspect outputs |
| Rows AI | Connected-source reporting, AI transformations and checkpoints | Moving to a different spreadsheet workflow; usage limits and source permissions |
The ten-minute accuracy test
- Copy a small, non-sensitive dataset with a known total, one duplicate row and a deliberately blank value.
- Ask for revenue by month, an outlier explanation and a formula for gross margin.
- Check the total yourself. Did the tool quietly drop blanks or count the duplicate twice?
- Change one source value and rerun. Does the answer update, or is it a static snapshot?
- Inspect formulas, edit history, sharing permissions and the export before using the result in a client report.
Bad fit if: you’re expecting any of these tools to certify your financial numbers. An elegant chart built from the wrong rows is still wrong. For source verification, use our research evidence guide and agent evaluation checklist.
Quick answers
Is the Excel =COPILOT() formula still supported? No. Microsoft says it was removed on 14 September 2026; use Copilot’s pane instead. Does Gemini cite sources in Sheets? Google announced citations for generated cells on 29 October, with rollout dependent on Workspace access. Can ChatGPT read Excel? Yes, supported XLSX/XLS/CSV files can be uploaded, subject to plan and file limits. Which is best for confidential data? Check the organisation’s data-handling policies, retention, permissions and approved tools before uploading anything.
More useful than another list of 50 AI tools: the differences that change what you can actually deliver. Continue with our AI presentations comparison, assistant chooser or free newsletter signup.
Source check: Google Workspace, 29 October 2026; Microsoft Excel COPILOT function notice; Copilot in Excel documentation; OpenAI data-analysis documentation; Rows AI Analyst documentation. Features vary by plan, region and rollout. No paid placements or hands-on accuracy scores claimed.
New vertical · 30 October 2026 · AI travel
Best AI trip planners: which ones can actually help you book?
An itinerary with lovely photos is not a hotel reservation. The useful distinction is inspiration vs live inventory vs checkout. Expedia’s 28 October release makes that difference worth examining again.
Start with Mindtrip for an ongoing assistant and itinerary; Expedia for activities and connected trip planning; Hotels.com for hotel trade-offs; Google AI Mode for supported US hotel booking and flight price tracking; Booking.com for accommodation search tied to its own booking flow. These are documented capabilities, not our hands-on booking tests.
Pick your fighter: what do you need done?
✈️ Expedia ↗Build activities around a trip and compare connected travel options.
🏨 Hotels.com ↗Compare shortlisted properties by location, room and policy.
🔎 Google AI Mode ↗Track flights; use hotel checkout where the US rollout supports it.
🧳 Booking.com ↗Search stays with AI and complete bookings in its marketplace.
Can it book, or only suggest?
| Tool | Useful AI capability | The booking boundary |
|---|---|---|
| Mindtrip | Preference-led itineraries, travel alerts, flight and stay choices | Says booking can happen directly or with assistant help; verify each supplier’s checkout |
| Expedia | Activity Planner and Property Expert | Eligible bookings through Expedia; availability varies |
| Hotels.com | AI Compare for shortlisted stays | Booking through Hotels.com; check room and cancellation terms |
| Google AI Mode | Flight price alerts and hotel comparisons | US English hotel booking rollout via integrated partners and Google Pay; merchant remains supplier |
| Booking.com | AI Trip Planner, property Q&A and trip support | Reservation completed in Booking.com; features vary by region |
The five-minute booking reality check
- Ask each tool for three bookable hotels for the same dates, guests and maximum total budget.
- Change one constraint: add a child, wheelchair access or a late arrival.
- Compare the quoted total with the real checkout price, including taxes and fees.
- Read the exact room type, cancellation deadline and payment merchant.
- Do not let an agent pay until you have approved the final details. A chat message saying “booked” is not a confirmation number.
Skip the magic trick: if an AI invents availability or can’t explain who takes your money, use it for inspiration, not transactions. Our agent evaluation checklist shows how to check real task outcomes.
FAQ: what changes in 2026?
Does Google AI Mode book flights? Google’s 27 August update covers flight price tracking and US hotel booking, not universal flight checkout. Can I use it outside the US? Some AI Mode travel features are available internationally, but Google described hotel checkout as a US English rollout; check your location. Best for families? Start with Expedia’s room and family-trip tools or Mindtrip’s preference-based planning, then verify occupancy and cancellation directly.
Explore our Gemini review, ChatGPT review, assistant comparisons and guides. Want a researched shortlist? The contact form records your interest but is not a newsletter subscription.
Research checked 30 October 2026 against Expedia’s Fall 2026 product release, Google’s AI Mode booking announcement, Mindtrip’s product documentation and Booking.com’s AI release. No hands-on transaction testing or paid placements claimed.
Assistants & personal agents
Compare assistants, research tools and personal agents by task, permissions and price. Includes AI shopping agents, browser permissions, privacy controls and checkout approvals. Start here to find the right fit.
Agents & automation
Autonomous agents, browser and coding agents, plus tools for monitoring actions, evaluating results and catching failures before deployment. Agent passports, identity and permissions → · Shopping agents & WebMCP → · Walmart agent spending limits → · Duet vs Dots vs Muse → · How to test an agent → · October agent safety incidents →
Creative AI
Image generation, design, emoji, video, filmmaking, avatars, dubbing, VFX, audio, music and voice.
Agentic video understanding
AI agents that navigate video timelines, inspect the right moments and turn video intelligence into operational workflows.
Productivity
Google Docs, Microsoft 365, Notion, email, meetings, transcription, notes, presentations and knowledge management. AI presentations: Gamma vs Pitch vs Beautiful.ai vs Canva → · Compare AI voice dictation: Wispr, Superwhisper, Aqua & Willow → · Compare bot-free AI meeting notetakers → · Best AI meeting follow-up tools → · AI for financial advisers → · AI research & evidence tools →
Legal AI →
Research, drafting, contracts and litigation. The useful tools now live inside real legal workflows — not just another chat window.
Developer AI
Coding copilots, IDE tools, debugging, testing, DevOps, APIs, app builders, databases and no-code platforms.
Insurance AI agents →
Pricing, underwriting, claims and servicing agents. Earnix’s new Agent Hub is a sign this vertical is moving from chat into real insurance decisions.
AI voice agents & phone calls →
Retell, Vapi, ElevenLabs and Rime Prism: real call workflows, interruptions, handoffs and permission checks.
Messaging & channels
WhatsApp, Telegram, Instagram, Messenger, Snapchat, Discord, Slack, Microsoft Teams and website chatbots. AI advertising and conversational agents →
Legal AI is moving from drafting assistant to workflow layer.
The category now spans source-grounded legal research, drafting, contract review, discovery, chronology building and firm knowledge. OpenAI’s 17 September 2026 launch of Astra for Law is another signal that frontier-model companies are moving directly into legal workflows and integrations. The test is not whether an AI sounds like a lawyer; it is whether a lawyer can trace the source, preserve matter permissions and review what the system did.
Prioritise authoritative sources, citations and jurisdiction coverage over polished prose.
First-pass research, document comparison, chronology, summarisation and drafting from verified material.
Unsupervised advice, fabricated authorities or workflows with no review trail. Hard pass.
AI agents are moving into underwriting, pricing and claims.
The useful distinction is simple: an insurance chatbot explains things; an insurance agent can gather context, analyse data, move work between systems and recommend or execute a next step. Earnix launched Agent Hub this week with 25+ insurance-specific agents, making this a category worth tracking now.
One to watch for insurance decisioning and orchestration.
Repetitive, data-heavy workflows with clear controls and measurable outcomes.
Opaque automated decisions with no human exception path. Hard pass.
Best AI voice agents for real phone calls: Retell vs Vapi vs ElevenLabs vs Rime Prism
A voice that sounds human is the easy demo. The hard part is whether it hears a correction, looks up the right order, asks before charging a card and hands off when it gets stuck. Voice quality is not task success. Here is the useful way to separate platforms from the speech models underneath them.
Pick your fighter: what are you building?
🧩 Vapi →Custom voice stack, API tools and multi-agent routing
🎙 ElevenLabs →Voice-first agents for web, phone and messaging
⚡ Rime Prism →Speech-to-speech model with text inspection before audio
The comparison most round-ups miss
| Product | What you buy | Best first test | Watch out for |
|---|---|---|---|
| Retell | Phone-agent platform | Transfer an upset caller with context | Actual total per-call cost |
| Vapi | Configurable orchestration platform | Change a booking mid-call | Component costs and engineering time |
| ElevenLabs Agents | Voice-first agent builder | Carry context from web to phone | Plan, language and integration eligibility |
| Rime Prism | Speech-to-speech model, not a complete call centre | Block an unapproved action before speech | Integration, language and beta fine-tuning |
Retell AI: phone operations first
Retell documents visual call flows, appointment booking, warm transfers, simulations, call monitoring and QA. Great if you need a receptionist or support line your team can operate. Bad fit if you only want a raw speech model to embed in your own infrastructure. Retell documentation ↗ · Compare Vapi
Vapi: control the plumbing
Vapi lets builders choose speech recognition, language models and voices, connect tools and route calls between specialised assistants. Great if you have engineering capacity and want flexibility. Skip it if maintaining multiple model and telephony components is the last thing your team needs. Vapi documentation ↗ · Compare Retell
ElevenLabs Agents: voice across channels
ElevenLabs documents visual workflows, API tools, knowledge bases, phone integration and web/mobile SDKs. Great if voice experience matters across multiple customer touchpoints. Bad fit if you cannot verify your required region, channel and compliance configuration. ElevenLabs documentation ↗ · Compare Prism
Rime Prism: new on 28 October
Prism is a speech-to-speech model with an inspectable text checkpoint before it speaks and tool calls alongside audio. Rime reports roughly 500ms response time and a stronger order-approval rate in a customer trial; those are vendor-reported, not independently reproduced. Fine-tuning APIs are being developed with beta customers. Great if you build your own voice stack and need guardrails over spoken output. Skip it if you want a ready-made phone support operation. 28 October announcement ↗ · Technical docs ↗
Five calls before you buy anything
- Interrupt the agent mid-sentence and correct a name or booking date.
- Give a price limit, then request an expensive alternative; confirm it asks before committing.
- Ask a question outside the knowledge base and check whether it admits uncertainty.
- Request a human handoff and inspect what context arrives with the transfer.
- Check call logs, audio retention, consent, real provider bills and end-to-end latency — not just model response speed.
FAQ: Is speech-to-speech automatically better? No: low latency and natural interruptions are useful, but task accuracy and permission controls still decide success. Is Rime Prism a direct substitute for Retell? No: one is a model and the other is an agent platform. Does a convincing voice mean the call succeeded? Absolutely not. See our agent evaluation checklist, voice dictation guide and meeting follow-up comparison.
Tell us which voice and agent topics you want covered. This is an interest enquiry, not an automatic newsletter subscription.
Sources checked 28 October 2026. Independent editorial selection based on public vendor documentation. No hands-on phone testing, paid placements or affiliate relationships claimed. Our methodology.
What every company profile will track
Beyond product features, profiles are designed to capture the business context that helps readers understand who is winning and why.
Start with our AI chatbot reviews, comparisons and guides. The directory will expand continuously as new categories and companies emerge.