Every CRM vendor is an "AI-powered platform" now. The phrase has become meaningless. What has not become meaningless is whether the features actually work, whether you can use them today, and how much they cost on top of your existing subscription.

We reviewed AI features across 15 CRM and customer service platforms — not based on press releases, but on what is generally available, what is gated behind enterprise pricing, and what is still sitting on a roadmap slide. The results are uneven. Some vendors have shipped real, measurable AI capabilities. Others have rebranded existing automation with an "AI" label and called it a product launch.

If you want the short version, here it is: the vendors charging per resolution (Zendesk, Intercom) are the ones most confident their AI actually works. Everyone else is hedging with credits, add-ons, or vague "AI-enhanced" language.

The four categories of CRM AI in 2026

Before the vendor breakdown, it helps to know what "AI in CRM" actually means right now. Nearly every feature falls into one of four buckets:

1. Copilots and drafting assistants. AI that writes emails, summarizes calls, or suggests replies. This is table stakes — almost every vendor ships some version of this. Quality varies, but the concept is universal.

2. Lead scoring and predictions. AI that scores leads, predicts deal outcomes, or flags at-risk accounts. Most of these use internal CRM data, which means they are only as good as what your team logs. Garbage in, garbage out.

3. Autonomous agents. AI that resolves customer tickets, qualifies leads, or books meetings without human involvement. This is the 2026 narrative — the shift from AI that helps to AI that acts. Only a handful of vendors have shipped agents that work well enough to charge per resolution.

4. Data enrichment and research. AI that fills in contact details, company information, or competitive intelligence from external sources. Useful, but often sold as a separate credit-based add-on that inflates your bill.

Vendor-by-vendor breakdown

HubSpot — Breeze AI

HubSpot has gone all-in on AI agents. The Spring 2026 Spotlight introduced a rebuilt Prospecting Agent, expanded Customer Agent controls, and a new outcome-based pricing model.

What shipped:

  • Customer Agent resolves support tickets across nine channels including web, SMS, WhatsApp, Instagram, and Slack. The Spring 2026 update added tone and style settings per channel, multi-brand support, and configurable working hours.
  • Prospecting Agent detects buying signals, maps buying committees, and drafts outreach for rep review.
  • Smart Deal Progression drafts follow-ups after calls and suggests CRM property updates based on deal history.
  • Breeze Assistant is the general-purpose copilot — content generation, email drafting, data queries.

What it costs:

This is where HubSpot gets complicated. Most Breeze features require Professional ($800/month) or Enterprise ($4,700/month for the Customer Platform). On top of that, AI runs on credits: $10 per 1,000 credits, with Customer Agent consuming 100 credits per conversation and Prospecting Agent using 10 credits per research task. As of April 2026, HubSpot also offers outcome-based pricing — you only pay when agents successfully complete tasks. There is a mandatory onboarding fee of $3,000 (Professional) or $7,000 (Enterprise). Free and Starter plans get minimal AI access.

The honest take: HubSpot is shipping real features, but the pricing excludes most small businesses. If you are on a Starter plan, Breeze AI is not meaningfully available to you.

Salesforce — Agentforce

Salesforce rebranded Einstein to Agentforce and is pushing autonomous AI agents as the centerpiece of its 2026 strategy. 12,000+ customers are live on Agentforce, and Salesforce reports 30-50% reductions in manual task time.

What shipped:

  • Agentforce agents reason, plan, and execute multi-step workflows using the Atlas Reasoning Engine.
  • Agent Builder is a low-code tool for customizing agents using Flows, Prompts, Apex, and MuleSoft APIs.
  • Agentforce Assistant (formerly Einstein Copilot) handles conversational queries inside the Salesforce interface.
  • Einstein Trust Layer provides guardrails for data privacy and accuracy across all AI actions.
  • Spring 2026 brought Agentforce 360 GA, Sales Workspace, two-way messaging, and a ChatGPT integration.

What it costs:

Three pricing models, which is itself a problem. The Conversations model charges $2 per conversation (any interaction within 24 hours). Flex Credits cost $500 per 100,000 credits, with standard actions at 20 credits ($0.10) and voice actions at 30 credits ($0.15). Per-user licensing is also available. You cannot mix Flex Credits and Conversations in the same org.

The honest take: Salesforce has the deepest enterprise AI capabilities, but the pricing complexity is a red flag. Most organizations still need consultants to implement Agentforce properly. If you are not already a Salesforce customer, this is not where you start.

Attio — AI-native from the start

Attio did not bolt AI onto an existing product. It built AI into the data model, which makes their implementation feel different from most competitors.

What shipped:

  • AI Attributes are custom fields that auto-fill using AI. They can summarize records, classify contacts into ICP tiers, or run a web research agent that searches external sources for funding stage, headcount, and more.
  • AI Workflows use AI-powered triggers for multi-step automations — not rigid if/then trees.
  • Automatic data enrichment pulls company and contact data from external sources and maps org hierarchies.

What it costs:

Free plan with 3 seats includes basic enrichment. Plus is $29/user/month, Pro is $69/user/month with call intelligence and sequences, Enterprise is $119/user/month. The AI Research Agent consumes 10 workspace credits per run, with additional credits at $70/month for 5,000 credits.

The honest take: Attio's AI feels native rather than bolted on, which is a genuine advantage for teams that want AI woven into their workflows. The credit costs for AI research can add up for heavy users, and the platform is still maturing — it hits real limits at scale. Best for teams of 5-50 who want modern tooling without Salesforce complexity.

Pipedrive — practical AI for pipeline management

Pipedrive has taken a quieter approach to AI. No autonomous agents, no outcome-based pricing. Instead, it focuses on features that reduce deal management friction.

What shipped:

  • AI Sales Assistant analyzes deals, predicts win probability, recommends next steps, flags at-risk deals, and surfaces overlooked higher-probability opportunities.
  • AI Email Writer generates drafts that match each rep's writing style.
  • AI-powered report generation lets you build reports from natural language prompts. Available on Premium and Ultimate plans.

What it costs:

Lite is $14/user/month with basic AI insights. Growth is $39, Premium is $49 with AI reports and advanced analytics, Ultimate removes feature limits. No free plan.

The honest take: Pipedrive's AI is not flashy, but it is practical. The Sales Assistant surfaces genuinely useful deal insights without requiring a consulting engagement to set up. The downside: Lite is too limited (no workflow automation, no email sync, no sequences), and the add-ons for campaigns ($16/month) and web visitors ($49/month) inflate the real cost.

Close — AI built for phone-heavy teams

Close has focused its AI efforts where its users spend the most time: on the phone.

What shipped:

  • Call Assistant records, transcribes, and summarizes calls with next steps and action items. Generates summaries in the language spoken on the call.
  • Predictive Dialer calls multiple leads simultaneously and connects reps only when someone answers. AI call scoring was recently added to rate and flag calls for coaching.
  • AI Email Draft Assist suggests follow-ups when email reminders come due.
  • AI Lead Summaries and AI Enrich auto-compile prospect information.
  • Notetaker for Meetings (public beta) extends transcription to video meetings — free for all Close customers.

What it costs:

Solo is $9/user/month. Growth is $99/user/month and is the first tier with AI email rewrite, AI lead summaries, and workflow automation. Scale is $139/user/month. Phone-related add-ons (call assistant, credits, premium lines) add 40-75% to the base price.

The honest take: Close is excellent for inside sales teams doing high-volume calling. The AI call scoring is a real differentiator. But the $99/month jump to access AI features is steep, and if you need visitor identification, chatbots, or marketing automation, you will be adding $500-2,000/month in third-party tools.

Monday CRM — AI agents for lead generation

Monday is newer to the CRM space, and its AI strategy reflects that — it is leaning heavily on AI agents to compensate for a less mature feature set.

What shipped:

  • AI Sales Agents (like "Lexi") autonomously source leads, enrich data, and engage prospects while warm.
  • AI Blocks are no-code building blocks for custom automations: sentiment analysis, auto task generation, email summaries. Each action costs 8 credits.
  • Deal Facilitator tracks deal progression, anticipates delays, and suggests actions.
  • AI Email Composer drafts outreach based on deal context and contact history.

What it costs:

Standard, Pro, and Enterprise plans get 500 free AI credits per month. Each AI Block action costs 8 credits per item, capped at one charge per item per 24 hours. CRM plans start around $12/seat/month.

The honest take: 500 free credits per month sounds generous until you realize each AI Block action costs 8 credits. That is 62 actions per month, which an active team burns through in a few days. Monday's AI agents are early-stage — good for SDR-type lead gen, but the CRM itself is less mature than dedicated platforms.

Freshworks — the best value play

Freshsales and Freshdesk run on Freddy AI, and the story here is value. Freshworks includes AI in its plans at price points that undercut everyone else.

What shipped:

  • Freddy AI Agents resolve up to 80% of customer queries without human intervention, with vertical-specific workflows for eCommerce, travel, logistics, and fintech.
  • Freddy AI Lead Scoring scores contacts based on engagement, profile fit, and activity patterns.
  • Freddy Copilot includes a writing assistant, reply suggester, and claims a 60% productivity increase.
  • Freshdesk's AI Agent Copilot suggests responses, summarizes tickets, and routes complex issues.
  • Behavioral signal detection reads website visitor behavior (rage clicking, confused scrolling) and alerts support teams.

What it costs:

Free plan for up to 3 users. Growth is $9/user/month with basic Freddy AI lead scoring. Pro is $39/user/month with full Freddy AI including deal insights and next-best-action. Enterprise is $59/user/month. No separate AI add-on fee.

The honest take: Freshworks is the most accessible entry point for CRM AI. Full AI at $39/user/month, no credits, no per-resolution charges. The trade-off is a smaller ecosystem, a less polished UI than newer entrants, and less brand recognition. For teams that want real AI without a four-figure monthly bill, Freshworks deserves serious consideration.

Zoho CRM — Zia AI at scale

Zoho has quietly built Zia AI across its entire 100+ product ecosystem. The scale is impressive even if individual features are less flashy than competitors.

What shipped:

  • Zia Agents (2025-2026) shift from AI that assists to AI that acts, built on Zoho's proprietary Zia LLM (1.3B to 7B parameter models).
  • Predictive lead scoring identifies prospects most likely to convert.
  • Churn prediction flags which customers might leave and from which product.
  • Natural language workflow creation — describe what you want in plain English, Zia builds the workflow.
  • Canvas view generation from images — upload a design mockup, Zia replicates it as a CRM template.

What it costs:

Free for up to 3 users. Standard is $14/user/month with basic Zia. Enterprise is $40/user/month with predictions, anomaly detection, and advanced analytics. Ultimate is $52/user/month with Zia Vision. No separate AI fees.

The honest take: Zoho's AI is genuinely capable, especially at the Enterprise tier. The concern is that Zia's proprietary models (1.3B-7B parameters) are much smaller than the GPT-5 and Claude models competitors use, which raises questions about quality ceiling for complex tasks. The breadth of the Zoho ecosystem is both its strength (everything connects) and its weakness (it can be overwhelming to configure).

Zendesk — the outcome-based pricing pioneer

Zendesk made the biggest structural move of 2026: in April, it removed the distinction between Advanced and Essential AI agents. All plans now get the full AI feature set, including agentic reasoning and API integrations. The pricing model says everything about their confidence.

What shipped:

  • AI Agents autonomously resolve inquiries across email, web, mobile, social, and voice.
  • Voice AI Agents handle phone calls with natural conversations.
  • Agentic reasoning enables multi-step procedures and external API calls — now included in all plans.
  • 3-click deployment — launch an AI agent by connecting knowledge sources, no scripting required.
  • Forethought acquisition (completed 2026) brings self-improving, self-learning AI agents.

What it costs:

Outcome-based: $1.50 per automated resolution (committed volume) or $2.00 per resolution (standard). Plans include free resolutions: Enterprise gets 15 per agent per month, Professional and Growth get 10, Team gets 5. Base Suite plans start at $19/agent/month.

The honest take: Zendesk charging per resolution is a statement of confidence. They only get paid when AI actually solves the problem. The April 2026 packaging change — making agentic reasoning available on all plans — removes the most common upsell friction. For customer service teams, this is one of the most mature AI implementations available.

Intercom — Fin AI and the $0.99 resolution

Intercom has the most talked-about pricing model in CRM AI: $0.99 per resolution. Fin AI Agent averages a 51% resolution rate across customers, up from 41% after 20+ feature upgrades.

What shipped:

  • Fin AI Agent resolves inquiries using RAG over your knowledge base, in 45 languages.
  • Fin Voice answers phone calls, handles complex questions, and hands off to humans.
  • Fin on Slack and Discord extends AI support to community channels.
  • Intercom Copilot enables human-AI collaboration — agents using it close 31% more conversations daily.
  • Topics Explorer groups conversations into topics for trend detection without manual tagging.

What it costs:

$0.99 per resolution — you only pay when Fin resolves the issue. One charge per conversation even with multiple follow-up questions. Requires an Intercom seat: Essential $29/seat/month, Advanced $85/seat/month, Expert $132/seat/month.

The honest take: Fin is one of the most mature AI agent products on the market. The 51% resolution rate is a real, published number — not a "up to" claim. The $0.99/resolution model means Intercom has skin in the game. The catch is the seat cost: $85/seat/month on Advanced adds up fast for larger teams.

The mid-market: who else is shipping real AI

A few smaller vendors deserve mention for AI features that are genuinely shipped, not just announced:

Folk — The folkX Chrome extension captures LinkedIn profiles with 91% enrichment accuracy. AI Research Notes powered by Perplexity generate contact briefs from public data. Follow-up Assistant detects stale conversations and suggests re-engagement. Premium plan at $48/user/month.

Nutshell — Timeline summarization, meeting transcription, and a Model Context Protocol (MCP) server that connects CRM data directly to ChatGPT and Claude. The MCP server is a notable differentiator. Business plan at $59/user/month.

Capsule — AI contact summaries, pipeline generator, and content assistant. The approach is deliberately incremental — reducing friction rather than flashy automation. Growth plan at $36/user/month.

Help Scout — AI Answers averages a 73% resolution rate in the Beacon widget. AI drafts, grammar checking, translation. All included in plans starting at $50/month. No per-resolution charges.

Copper — Copper GPT for natural language CRM queries, AI email rewriting, and LinkedIn email finder (beta). Professional plan at $59/user/month.

The pricing landscape

AI pricing in CRM has fragmented into four models, and understanding which one your vendor uses matters more than the feature list.

Pricing model How it works Who uses it What it signals
Per resolution Pay only when AI resolves an issue Zendesk ($1.50-2.00), Intercom ($0.99) High confidence in AI quality
Credit-based Buy credit packs, features consume credits HubSpot, Salesforce, Attio, Monday Hedging — vendor is not sure how much AI will be used
Included in plan AI bundled at each tier Freshworks, Zoho, Pipedrive, Help Scout Simplest to budget, but advanced features are tier-gated
Outcome-based Pay when agents complete tasks HubSpot (new, April 2026) Experimental — too early to evaluate

The per-resolution vendors are putting skin in the game. If their AI does not work, they do not get paid. Credit-based pricing shifts the risk to you — you pay for AI actions whether or not they produce useful results.

What is real vs. what is marketing

Three signals separate vendors with real AI from vendors with an AI label:

1. Published resolution or success rates. Intercom publishes 51%. Help Scout publishes 73%. Zendesk's outcome-based pricing implies measurable results. Vendors that do not publish numbers are telling you something.

2. Outcome-based or per-resolution pricing. If a vendor charges per resolution, they are confident the AI resolves issues. If they charge per credit or per action, they are charging for usage regardless of outcome.

3. Specificity of feature descriptions. Compare "AI-enhanced workflows" (vague, could mean anything) with "AI agent that reads, categorizes, and responds to support tickets, escalating complex issues to humans" (specific, testable). The vague ones are usually automation with an AI sticker.

Gartner predicts 40% or more of agentic AI projects will be scrapped by 2027 — not because the models fail, but because organizations cannot operationalize them. The CRM vendors that make AI easy to deploy (Zendesk's 3-click setup, Freshworks's included-in-plan approach) are better positioned than those requiring consultants and complex configuration.

The bottom line

The CRM AI landscape in 2026 is genuinely bifurcated. On one side: Zendesk, Intercom, and Freshworks have shipped AI that works well enough to measure, price transparently, and deploy without a consulting engagement. On the other: Salesforce and HubSpot have the deepest capabilities but behind pricing and complexity barriers that put them out of reach for most small and mid-size teams.

The mid-market — Attio, Pipedrive, Close, Folk — is where the most interesting work is happening. These vendors are shipping focused, practical AI features that solve specific problems (call scoring, LinkedIn enrichment, deal predictions) rather than trying to build autonomous everything.

The question is not "which CRM has AI" — they all do. The question is whether the AI your vendor shipped actually reduces the work your team does today, and whether you can afford it at your current seat count. Start there.