Mike HeapMike is an experienced Product Manager who focuses on all the “non-development” areas of My AskAI, from finance and customer success to product design, copywriting, testing and more.
Table of Contents
- Ada CX vs Forethought at a glance
- How are Ada CX and Forethought different from their old 'AI bots'?
- How does an AI customer service agent work in Ada CX and Forethought?
- How can I use my AI agent in Ada CX and Forethought?
- Direct replies
- Copilot replies
- Voice
- Which is easier to set up: Ada CX or Forethought?
- How do Ada CX and Forethought differ in what they can be trained on?
- 'Static' content
- 'Dynamic' content
- Which has better answer quality, Ada CX or Forethought?
- Is Ada CX or Forethought easier to improve?
- Which has more features: Ada CX or Forethought?
- How easy is it to customize Ada CX and Forethought?
- What about vendor lock-in?
- Does Ada CX or Forethought have any other AI features?
- Tagging
- Agent translation
- Sentiment / CSAT prediction
- What about security, is Forethought more secure than Ada CX?
- Which costs more, Ada CX or Forethought?
- The pricing model
- The overall cost
- Conclusion - should I choose Ada CX or Forethought?
- FAQs
Created time
Aug 25, 2026 09:15 PM
Title length (<60)
Author
Last optimised
Ecomm?
Tags
Image
Publish date
Aug 6, 2026
Video
Slug
ada-cx-forethought-comparison-2026
Featured
Type
Article
Ready to Publish
Ready to Publish
Ada CX is an agentic customer service platform you buy and implement through Ada's own team. Forethought is a Zendesk-owned multi-agent suite that still runs on rival helpdesks. Choose Ada CX for native social channels and a downloadable compliance pack. Choose Forethought for helpdesk reach, ticket triage and automated QA in one contract.
You are weighing up two AI support platforms, and the one owned by a helpdesk vendor turns out to reach more helpdesks: Ada's documentation enumerates 12 handoffs, against the 23 helpdesks and CRMs Zendesk-owned Forethought publishes. I assumed the opposite. Neither vendor's own pricing page carries a dollar figure, and Vendr's procurement records put the medians about $1,000 apart, at $73,500 a year for Ada and $74,483 for Forethought.
Ada CX and Forethought both sell an agentic AI layer for customer support, and they arrive at it from opposite directions. Ada is independent, sales-led and implemented by its own professional-services team (there is no public signup and no published price). Forethought started out independent in 2017 and now belongs to Zendesk, which closed the acquisition in March 2026 and put the product on sale that June.
The teams choosing between them are usually enterprise CX buyers with a helpdesk already in place, and that helpdesk tends to settle it. Ada's own documentation names 12 support platforms it can hand off to, and neither Intercom nor HubSpot nor Freshdesk appears on that list. Forethought, the one owned by a helpdesk vendor, publishes 23.
Neither one puts a number on its pricing page, so Vendr's procurement records are the only observed contract data in this comparison, and they land the two medians within about $1,000 of each other. We scored both across eleven categories on published evidence alone, and Forethought came out ahead on reach and tooling while Ada held channel breadth and security.
Ada CX vs Forethought at a glance
Ada CX | Forethought | |
What it is | Agentic CX platform, sold direct by Ada | Zendesk-owned multi-agent AI suite |
Headline price | No figure on its pricing page; $73.5K median/yr (Vendr, n=112) | No figure on its pricing page; $74.5K median/yr (Vendr, n=69) |
How you buy it | Sales-only; your Ada rep hands you a demo agent link | Sales-only; Zendesk's help center says contact Zendesk Sales |
Best for | Voice and contact-center-heavy enterprise CX | Teams on a non-Zendesk helpdesk wanting Triage and QA |
Not a fit for | Intercom or HubSpot shops | Teams under 20,000 historical tickets |
Biggest limitation | 12 documented handoffs, no Intercom or HubSpot | 20,000-ticket floor, 30 to 90 day implementation |
Scorecard | 2 wins of 11, 4 tied | 5 wins of 11, 4 tied |
How are Ada CX and Forethought different from their old 'AI bots'?
TL;DR: Both now sell a multi-agent suite rather than a chatbot. Ada runs one reasoning layer across chat, voice, email and social. Forethought splits the job across five named agents, with Autoflows underneath.
Ada has raised about $192M across seven rounds, peaking at a $130M Series C at a $1.2B valuation in May 2021. You buy and implement it through Ada's own team, which is the practical half of the ownership question (who picks up the phone when the agent misbehaves).
Ada's current product is built on a reasoning engine, which Ada announces as its Unified Reasoning Engine. What its own docs show is one build reaching chat, voice, email and social, with Persona, Languages and Custom Instructions set once.
If you run more than two channels, make that single layer the first thing you stress-test.
Forethought was founded in 2017 by Deon Nicholas and Sami Ghoche in San Francisco, and raised $115M+ before its sale. Its homepage now names five agents: Discover, Solve, Triage, QA Agent and an Agentic AI Copilot (five products to learn). Autoflows, its logic layer, takes business rules in plain language and works out the steps itself.
Zendesk announced a definitive agreement on March 11, 2026 and closed on March 26, 2026. The product became available to buy on June 4, 2026, and you buy it from Zendesk Sales (check which entity appears on your contract before you sign).
Both products review well on G2:
Ada CX | Forethought | |
G2 rating | 4.6 / 5 | 4.3 / 5 |
G2 page |
How does an AI customer service agent work in Ada CX and Forethought?
TL;DR: Both retrieve from your knowledge and ticket history, then act. Ada routes everything through one reasoning layer. Forethought splits the job across separate Solve, Triage, Discover and QA agents.
Ada runs a single reasoning layer, so you build once (the same logic then reaches chat, voice, email and social). Its four stated pillars are multi-channel deployment, the reasoning engine, Coaching plus Playbooks, and enterprise integrations with guardrails.
Forethought splits the work by job. Solve answers customers, Triage classifies and routes, Discover finds missing knowledge, and QA Agent scores the result.
Underneath sits SupportGPT, which pulls answers out of your own content and returns them in the 0.03 to 0.13 second band (fast enough that response speed will not be your problem).
Forethought hosts hundreds of per-customer fine-tuned models, meaning each one is trained on that customer's own data. Ada tunes behavior through configuration and coaching on top of one shared engine.
I read that as a straight trade-off: Ada's one layer holds behavior consistent on chat, voice, email and social, and Forethought's four surfaces buy per-job tooling for the cost of learning all four.
How can I use my AI agent in Ada CX and Forethought?
TL;DR: Both do direct replies and both do voice. Forethought adds a Chrome-extension copilot and a headless API. Ada reaches WhatsApp, Messenger and Instagram natively, which Forethought only gets through Intercom.
Mode | Ada CX | Forethought |
Direct customer replies | ✅ | ✅ |
Agent-facing copilot | Not documented | ✅ (Chrome extension) |
Voice | ✅ | ✅ (Professional tier) |
Social (WhatsApp, Messenger, Instagram) | ✅ | ✅ (via Intercom) |
Slack | Not listed | ✅ (Professional tier) |
Headless API | ✅ (custom channels) | ✅ (Enterprise tier) |
Direct replies
Direct replies are the primary mode on both. Ada's documented channel set is chat, voice, email, social and third-party. Forethought's Solve covers chat, email, voice, SMS, Slack and mobile.
Forethought's channels are tier-gated, and the tiers were repackaged this year (check your quote against this year's tier list). Chat and mobile come with Team, email, voice and Slack sit on Professional, and the Solve API is Enterprise only.
Copilot replies
Forethought's Assist, now presented as the Agentic AI Copilot, is a Chrome browser extension that overlays your helpdesk. That is a real deployment constraint, since every agent has to install the extension and then work inside a browser overlay.
Ada's documented surfaces are customer-facing. Its escalation path is handoff integrations into Zendesk, Salesforce, ServiceNow, Gorgias, Freshchat, Help Scout and Kustomer among others.
Voice
Both do voice. Ada's second-generation voice integrates with NICE CXone, Genesys Cloud and PureConnect, Amazon Connect, Aircall and Twilio Flex (six contact-center backends, which is the deepest list here). Forethought launched Voice AI in March 2025 with Cartesia, on the same Autoflows engine as chat and email.
If you sell to consumers, put social near the top of your scorecard.
Ada reaches WhatsApp, Facebook Messenger and Instagram Direct through Sunshine Conversations. Its social channel docs draw the ceiling: "Ada does not support using other tools than the ones listed above with Sunshine Conversations."
Forethought reaches the same three networks through its Intercom integration, so social depends on you running Intercom.
Which is easier to set up: Ada CX or Forethought?
TL;DR: Neither is quick. Forethought quotes 30 to 90 days and wants 20,000+ historical tickets. Ada runs 8 to 16 weeks with a CSM you keep depending on afterwards.
Forethought publishes its own prerequisites. From its FAQ: "Forethought's AI performs effectively with a historical ticket data volume of 20,000+ and requires at least 2,000 email or chat tickets per month to operate smoothly."
The timeline comes from the same page:
"Most teams are fully set up within 30–90 days, and once implementation is complete, you'll be paired with a Customer Success Manager to support ongoing success."
Word-on-the-street runs faster than that.
"We really like Forethought's flexibility to build highly customized workflows that suit our business needs. The initial setup was very easy, which we appreciated." — Lea R., mid-market reviewer, April 2026, G2
Grammarly reported implementation in about a week and a half.
Ada is sales-led with no public signup. The published path is a consultation, then discovery, then a quote, then a professional-services implementation.
You will want the implementation scope in writing at the quote stage, because that is where the weeks go (scope is what stretches a timeline).
A full enterprise deployment runs 8 to 16 weeks. Chat-only knowledge-base pilots can go in 4 to 6 weeks, and regulated voice deployments across several backends stretch toward 4 to 6 months.
Ada's own getting-started docs are candid about who does the work: "Most people who obtain access to their AI Agent don't go through this setup process. Ada's Customer Experience (CX) team usually configures AI Agents in advance, adding Knowledge (at minimum), and sometimes also customizing the Persona."
Both let you test before going live. Pin down the scope of each test suite before you rely on it (the two suites do not test the same thing).
Ada runs Simulations over multi-turn test cases you author. Forethought's Triage LLM Simulations replay real historical tickets (LLM meaning the model that reads a ticket and picks its category), capped at 1,000 per run and visible to Triage Admins only, and they cover the classification model.
Solve's customer-facing answers get a preview pane.
Each suite stays inside its own scope: Ada's on the test cases you author, and Forethought's on the classification model, so your own past customer conversations stay untested end to end.
If you would rather not spend a quarter on implementation, My AskAI connects to the helpdesk you already run, and we go live the same day with no historical-ticket prerequisite. We install into Zendesk, Intercom, Freshdesk, Freshchat, Gorgias and HubSpot.
How do Ada CX and Forethought differ in what they can be trained on?
TL;DR: Forethought ingests the modern knowledge stack, including Notion, Confluence, Google Drive, SharePoint and Shopify. Ada ingests formal help centers only, with no Notion, no Confluence, no Drive and no SharePoint.
Source | Ada CX | Forethought |
Help center articles | ✅ | ✅ |
Past tickets | ✅ | ✅ (20,000+ recommended) |
Notion | ❌ | ✅ |
Confluence | ❌ | ✅ |
Google Drive | ❌ | ✅ |
SharePoint / OneDrive | ❌ | ✅ |
Salesforce | ✅ (Knowledge) | ✅ |
Shopify | ❌ | ✅ |
MCP | ✅ (admin and authoring server) | ✅ (knowledge connector) |
Internal-only marking | Not documented | Not documented |
'Static' content
Ada connects to formal, structured knowledge systems. Its documented support-platform handoffs number 12: Salesforce, Zendesk, Amazon Connect, ServiceNow, Genesys, Gorgias, Freshchat, Help Scout, Kustomer, Twilio Flex, Gladly and Dixa. Its knowledge-base importers number 8: Zendesk Guide, Salesforce Knowledge, Contentful, Dixa Knowledge, Gladly, ServiceNow Knowledge Management, Helpjuice and Microsoft Dynamics.
As of August 25, 2026, that list carries no Notion, no Confluence, no Google Drive, no SharePoint or OneDrive, no Slack and no Shopify. Ada's knowledge-integration docs have an answer for anything outside the list:
"If you use different authoring tools, you can contact your Ada team, so they can use a custom integration that collects information in your knowledge base and uploads it into your AI Agent using the Knowledge API."
That is a services ticket. Price it into the quote before you sign (custom-integration work is where the scope creeps).
Forethought lists 18 knowledge and LMS systems, Notion and Confluence among them, alongside Coda, Contentful, Document360, eGain, Glean, Guru, Helpjuice, ReadMe, Shelf and Stonly.
'Dynamic' content
Forethought's connector list adds Google Drive, SharePoint, Shopify, Airtable, Amazon S3, Snowflake, WordPress, Linear, Canny and an MCP connector. Its primary training inputs are historical tickets, help-center articles, past macros and canned responses, and CRM data.
Both list MCP, for different jobs. Forethought lists MCP as a data connector on its integrations page. Ada runs an MCP Server for administration and authoring, covering bulk test-case creation, tool authoring through a change set, and CSV export (admin plumbing for your own team).
Ada's docs name 12 helpdesk handoffs. Forethought's own integrations page lists 23 helpdesks and CRMs, and the ones Ada never names include Intercom, HubSpot, Freshworks, Front, Gong, Jira and TeamSupport.
Those two counts are the first thing to check in your own evaluation.
If your helpdesk is Intercom and your documentation lives in Notion, Ada's lists cover neither. My AskAI runs natively inside Zendesk, Intercom, Freshdesk, Freshchat, Gorgias and HubSpot, and we connect to Notion, Google Drive, Confluence, SharePoint, OneDrive, Dropbox, Shopify and Salesforce knowledge pages.
Which has better answer quality, Ada CX or Forethought?
TL;DR: Both publish case-study bands in a similar range, Ada 45 to 84% and Forethought 60 to 87%. Neither number is independently audited or defined the same way, so read them as vendor claims.
Take both sets of numbers with a grain of salt: these are the wins each vendor chose to publish. Ada's published case studies run from 45% to 84%. Tilt reports 84% automated resolution on chat and Loop Earplugs 80% CSAT, both on that index.
Wave reports a 65% ticket reduction in month one and Betsson 45% interaction resolution on the same index, and BlueJeans by Verizon 72% containment.
Forethought's dated case studies run from 60% to 87% (dated: you can see how old each claim is). Grammarly reports 87% deflection with 4.2 CSAT.
YAZIO reports 80% deflection with CSAT rising from 3.1 to 4.0 over six months. On Forethought's customer index, YNAB reports 70% deflection, up from 25% with a previous chatbot, Kickfin 72% self-serve, and ActiveCampaign 60%+ chat deflection. Forethought's homepage also carries a ceiling figure of 98% on a resolution-rate stat tile, which reads to me as a best case.
Ada publishes automated resolution and containment. Forethought's figures are deflection and self-serve rate.
Almost every vendor in this category counts a deflected conversation and a resolved one differently, so the bands are not comparable to the decimal.
For context on where the field sits, My AskAI's AI resolution rate benchmark study covers 195 rated deployments across 38 vendors, and puts the field median at 70%. Use that as your baseline and read both vendors' case-study bands against it.
Both vendors' bands straddle that median.
Is Ada CX or Forethought easier to improve?
TL;DR: The two improve differently. Discover writes the missing articles, QA Agent scores every interaction, and Triage simulations test a routing model on real tickets. Ada regression-tests the answering agent itself, up to 3,000 simulations a day, with coaching applied straight from a failure.
Forethought's Discover Agent reads historical tickets and help-center content, finds the questions your knowledge cannot answer, and auto-generates articles and Autoflows to fill them. Article creation and gap detection sit on Enterprise, or arrive as an add-on lower down.
QA Agent is my favorite thing either vendor has built here.
It scores 100% of interactions on empathy, grammar, relevance and resolution quality, against a manual review sample that is usually under 5%. You get one rubric on Team, five on Professional and 20 on Enterprise.
Triage LLM Simulations test the routing model. Forethought's own help-center article, archived January 2025, describes them as a way to test a Triage model against real historical ticket data before it goes live. Each run takes the oldest 1,000 tickets in the range you select (the oldest ones, which matters if your ticket mix has moved).
Triage setup draws criticism too.
"Forethought Triage LLM Setup can be enhanced, and they can provide better training and support for it!" — Bojan S., enterprise reviewer, G2
Ada improves through Playbooks and Coaching. Playbooks encode the procedure you want followed. Coaching lets a reviewer leave granular feedback on a past conversation that the agent then learns from (one conversation at a time).
One support director rates the Playbooks loop:
"My favourite feature is the playbooks. You write out what you want, and Ada turns your thoughts into an effective protocol for your chatbot" — Danielle G., Director Customer Support, G2
Ada's Simulations test the answering agent the same way. You can run up to 3,000 simulations a day across up to 1,000 test cases, at up to 40 turns each, on chat, email and voice (that is a lot of headroom).
You can apply Coaching straight from a failed simulated conversation. One operational caveat before you run a batch:
"Actions are not mocked—they execute against live systems." — Ada, Simulations documentation
That one made me sit up. Ada also publishes dated release notes, with entries running through August 2026, so you can check what changed and when.
Which has more features: Ada CX or Forethought?
TL;DR: Forethought wins on breadth, with five named agents plus Browser Agents, a headless API and MCP. Ada's depth is in voice and contact-center routing, which Forethought reaches less deeply.
Capability | Ada CX | Forethought |
Named agent products | 1 reasoning engine | 5 (Solve, Triage, Discover, QA, Copilot) |
Contact-center routing | ✅ (NICE, Genesys, Amazon Connect, Twilio Flex) | Not documented |
Native social channels | ✅ | ❌ (via Intercom) |
Automated QA scoring | ❌ | ✅ |
Browser agents (no API needed) | ❌ | ✅ |
Headless deployment | ✅ (open APIs) | ✅ (Headless Forethought) |
Multi-brand | ✅ | ✅ (2 on Team, 20 on Professional) |
Image inputs | ✅ | ❌ |
Forethought's breadth shows in Solve: Autoflows, a Workflow Builder, a custom Action Builder against any public API, and built-in CSAT collection.
Browser Agents, launched October 2025, act inside browser-based systems that expose no API at all. Headless Forethought embeds the whole thing through APIs or MCP (no chat widget required).
Ada goes deep on voice: NICE CXone, Genesys Cloud and PureConnect, Amazon Connect, Aircall and Twilio Flex. Ada also documents handoffs into Salesforce Chat and Messaging, Zendesk Chat, Support, Ticketing and its SMTP connector, ServiceNow, Gorgias, Freshchat, Help Scout and Kustomer.
Around that sit Playbooks, Proactives, Greetings, Custom Instructions, Glossary, Variables, Personas, Redactions and the Simulations suite. That is a deep bench for one channel family. Weigh it against Forethought's five products before calling breadth the winner (depth on one channel family is a different bet from breadth across five).
Reviewers name limits on both sides. On Ada, an enterprise reviewer put the ceiling like this:
"Ada automates a large portion of our customer interactions... it can feel a bit limited when building more dynamic workflows or adapting conversation logic automatically based on user intent." — Enterprise reviewer, G2
On Forethought, reviewers name missing image support, analytics dashboards that read as cluttered, no live preview for workflow builders, and slow fixes on long-standing bugs.
Every one of those is worth putting on your demo agenda. Ask Forethought about image support first (Ada ships image inputs, Forethought does not).
How easy is it to customize Ada CX and Forethought?
TL;DR: Forethought's Autoflows take business logic in plain language, and its Action Builder hits any public API. Ada's Playbooks and MCP-authored change sets cover the same ground, though reviewers describe a ceiling on dynamic workflows.
Autoflows take a policy written in plain language and work out the steps. Forethought sold it that way itself on an AWS Marketplace listing, since pulled: a no-code agent builder for support, onboarding and returns.
The Action Builder is what turns that into work. If you demo one thing at Forethought, make it the Action Builder running against your own API (their demo data will always look clean).
It calls any public API, so a Shopify order lookup or a refund through your payment system is configuration work. Per-customer fine-tuned models sit underneath.
The learning curve is real, though.
"Forethought typically has a higher than expected learning curve, making it a bit challenging to get ramped up and know exactly how to implement it best for specific use cases." — Verified utilities reviewer, May 2025, G2
Ada's builder is also no-code, covering flows, knowledge sources, brand voice and handoff rules. Playbooks have a documented step reference: Action, Text, Link, Scheduled, Conditional, Handoff, List Option, Set Variable and Instruction blocks (nine block types, all documented). Custom Instructions, Glossary, Variables and Personas sit alongside.
Ada's newer authoring route goes through its MCP Server. I would ask for a demo of the change-set model, where behavior changes are staged, tested against simulations, released only after explicit approval, and reverted in one step.
The ceiling reviewers describe is the dynamic-workflow limit, plus the design time Playbooks need to work well.
What about vendor lock-in?
TL;DR: The helpdesk-owned product is the more portable one. Forethought is Zendesk-owned and still lists 23 helpdesks and CRMs, Intercom and HubSpot among them. Ada's own docs name 12 handoff integrations, and neither Intercom nor HubSpot is one of them.
I find Forethought's helpdesk reach where it was before Zendesk bought it.
Zendesk's own help center describes the product it now owns as standalone. Verbatim, from its article on Forethought AI agents by Zendesk: "Forethought AI agents by Zendesk is an AI-powered support platform that can operate independent of the Zendesk platform".
On buying it, from that same Zendesk help center article: "Forethought plans are available for purchase for any customer. Contact Zendesk Sales."
That commitment was made at the announcement on March 11, 2026, restated at general availability on June 4, 2026, and it stands in Zendesk's live documentation today (three statements, three months apart). The June 4 GA announcement worded it: "If you're not a Zendesk customer, Forethought is available to purchase and can be integrated with your platform and work across all supported channels."
The catalog backs it up. Five months after close, Forethought's integration catalog still names 23 helpdesks and CRMs under its own Helpdesks & CRM tab. That list runs Comm100, Drift, Dynamics365, Freshchat, Freshworks, Front, Gainsight, Gladly, Gong, Gorgias, HaloCRM, Help Scout, Hubspot, Intercom, Jira, Kustomer, NICE, Salesforce, ServiceNow, Talkdesk, TeamSupport, Zammad and Zendesk.
Ask Zendesk to put the non-Zendesk roadmap in writing at contract stage anyway (an independence promise in a help-center article is not a contract term).
Ada's handoff integration docs name 12 support platforms: Salesforce, Zendesk, Amazon Connect, ServiceNow, Genesys, Gorgias, Freshchat, Help Scout, Kustomer, Twilio Flex, Gladly and Dixa. Intercom, HubSpot and Freshdesk are not among them. Fun fact: the one Freshworks product Ada does name is Freshchat, which is an easy one to misread on a vendor call.
If your team runs one of those three, Ada is out before the evaluation starts. What happens to a trained agent when you change helpdesk is its own criterion, and both vendors leave it to the migration.
If you may change helpdesk inside the contract term, My AskAI installs into Zendesk, Intercom, Freshdesk, Freshchat, Gorgias and HubSpot, and our trained agent moves with you, so the custom answers, guidance and tasks carry over.
Does Ada CX or Forethought have any other AI features?
TL;DR: Forethought's Triage Agent classifies, routes and scores sentiment, and QA Agent grades every conversation. Ada handles multilingual natively, with no equivalent automated QA layer published.
Feature | Ada CX | Forethought |
Ticket tagging and classification | Not documented | ✅ (Triage Agent) |
Sentiment detection | Not documented | ✅ |
Spam filtering | Not documented | ✅ |
Automated QA scoring | ❌ | ✅ (QA Agent) |
Multilingual answers | ✅ | ✅ (Team tier and above) |
Manual translation overrides | ✅ | Not documented |
PII redaction controls | ✅ | ✅ (on by default) |
Tagging
Triage is Forethought's classification layer, and it auto-classifies, prioritizes and routes on sentiment, language, urgency, spam and product type. You get three models on Team, six plus custom handoff models on Professional, and Enterprise adds detection of missing knowledge and Autoflow coverage.
Ada's routing sits inside Playbooks and handoff rules.
If ticket tagging is on your requirements list, score it for Forethought and move on (Ada documents no equivalent).
Agent translation
Both handle multiple languages, and Forethought's packaging moved in the customer's favor this year. Multilingual sits on the entry Team tier after the 2026 repackaging. Solve chat auto-detects the language; Solve Email needs your CSM to switch it on.
Ada handles multilingual across every channel, with documented language switching, multilingual setup and manual translation overrides. Use those overrides for any legal wording that must not be paraphrased (refund policies, warranty terms, anything legal signed off on).
Sentiment / CSAT prediction
Forethought's QA Agent grades every conversation on empathy, grammar, relevance and resolution quality. Triage scores sentiment on the way in. Classification and automated quality scoring arrive from one vendor here, and that is the strongest single argument for Forethought.
Ada's improvement path is Coaching plus Simulations, which exercise the agent against test cases you author. Ada collects CSAT and has Redactions settings, and automated QA scoring stays with Forethought (ask Ada where that sits on its roadmap).
What about security, is Forethought more secure than Ada CX?
TL;DR: Ada publishes more, and publishes it openly, through a trust center with downloadable SOC 2 Type II, SOC 3, HIPAA, PCI AOC and pen-test reports. Forethought holds SOC 2 Type II and HIPAA, behind a customer-gated Trust Report.
Ada runs a public trust center, and it is the most complete compliance pack of the two (no sales call, no NDA).
The grid there is broad and downloadable: SOC 2 Type II and SOC 3 covering October 2023 to September 2024, HIPAA, a PCI AOC, GDPR, CCPA, CPRA and PIPEDA, plus 2025 penetration-test and AI Agent Security reports, data-flow diagrams and a VPAT.
Ada was also the first AI customer service platform to earn AIUC-1, the newer AI-specific standard. SecurityScorecard rates it A, as does Qualys.
Ada's gap is the older frameworks. Its AI-specific standard is AIUC-1, and the ISO 27001 and ISO 42001 rows are blank.
If your procurement template asks for ISO 27001 by name, flag that early with the procurement team (the row is simply blank).
Forethought holds SOC 2 Type II and HIPAA with annual audits, is GDPR and CCPA compliant, and aligns with the NIST Cybersecurity Framework and NIST 800-53. Data is AES-256 encrypted at rest, PII, PHI and financial data (personal data, health data and payment details) are redacted by default, and RBAC, SSO and customer-specific data segmentation are standard.
Its Trust Report gives 24/7 access to audit logs and security documentation, in-product and gated to customers.
Check two details before the paperwork starts.
First, on ISO 27001, Forethought's own security page draws the distinction: "Forethought is independently audited and certified to meet compliance standards for security, availability, and confidentiality. We are compliant with ISO 27001 and certified for SOC 2."
Its dedicated compliance page covers SOC 2, GDPR and CCPA, and never mentions ISO 27001 at all. The security policy explains where ISO enters: the SOC 2 Type II report carries control mappings to ISO 27001 (mappings, which is a control crosswalk).
Second, the acquisition did not transfer its parent's certifications. Zendesk publishes an in-scope product list per compliance program, updated July 7, 2026, and Forethought is named in none of them. That covers ISO 27001, 27017, 27018 and 27701, ISO 42001 and FedRAMP LI-SaaS.
The AI Agents entry on those lists is Zendesk's own native agents product, which is a separate SKU. Zendesk has published nothing extending any certification to Forethought.
Ask Forethought for its own certificates, since Zendesk's do not extend to it (the SOC 2 report itself, under NDA if that is what it takes).
Read the two together and the gap is narrow: Ada lets a security reviewer download more before signing, and Forethought gates its Trust Report to customers.
A certification covers the vendor's management system. Put your own guardrail configuration on the review list either way.
Which costs more, Ada CX or Forethought?
TL;DR: Neither vendor's own pricing page carries a dollar figure, a rate or a unit definition, and both route you to a quote. Vendr's procurement data puts them roughly $1,000 apart at the median, Ada at $73,500 a year and Forethought at $74,483. Neither offers a free trial, so you cannot check either number without a sales call.
The pricing model
Ada bills by conversation, which has one useful consequence: the bill is independent of how well the AI performs. A better resolution rate lowers your cost per resolved ticket without raising the invoice, which is what I want from a meter.
Vendr's procurement data puts overage at $0.10 to $0.50 per conversation, and it applies only above the volume allowance in your contract.
Ada argues the case publicly. From its pricing blog: "For most enterprises adopting AI customer service at scale, conversation-based pricing offers stronger alignment between performance, cost control, and long-term customer experience outcomes." It also declines to define a resolution.
Push on that definition in the call.
Ada leaves its own unit undefined, and its pricing page carries one concrete figure: "350+ businesses" across seven industries.
Forethought describes its model in one sentence on its pricing page: "Our pricing model is a blend of platform access fees and an outcome-based pricing cost." There are no dollar figures anywhere on that page, every button is a quote or demo request, and payment terms are annual. Since the acquisition, Zendesk's help center also tells buyers to compare Forethought plan summaries on the Zendesk pricing page, so ask for that comparison too.
Neither vendor defines its own meter where you can check it: Forethought's outcome-based model publishes no rate and no qualifying definition, and Ada's conversation-based model leaves "conversation" undefined.
Both meters need a quote before you can forecast them. Treat both headline models as marketing until you have one in hand.
The overall cost
Ada CX | Forethought | |
Figure on its own pricing page | ❌ None | ❌ None |
Pricing model | Conversation-based | Platform fee plus outcome-based cost |
Observed median (Vendr) | $74,483/yr, n=69 (page dated Feb 2026, pre-close) | |
Observed range | $36,026 to $318,400 | $35,670 to $151,400 |
Unit | Per conversation, overage above allowance | Per conversation observed by Vendr, $0.50 to $2.00 |
Free trial | ❌ Consultation booking | ❌ Sales-mediated Proof of Value |
Payment terms | Annual | Annual |
Vendr's procurement data is the only third-party number either buyer can see. Ada's marketplace page puts the median all-in contract at $73,500 a year across 112 purchases, and describes the model:
"Pricing is based on a combination of conversation volume, feature tier, and deployment complexity,"
At Vendr's median, the effective rate works out around $0.245 per conversation. That is the closest thing to a unit price I can give you for Ada. The same Vendr page bands first-year platform fees at roughly $30K to $70K for small deployments, $70K to $150K for mid-market and $150K to $300K+ for enterprise, each with implementation on top.
Forethought's marketplace page puts the median at $74,483 a year across 69 purchases, in a range of $35,670 to $151,400. That page is dated February 2026, which is before the Zendesk deal closed, so it describes the standalone company's contracts.
Vendr's observed Solve unit pricing sits in a $0.50 to $2.00 per conversation band (a four-fold spread, so your own volume decides where you land).
On predictability, Forethought's fixed annual contract fixes your cost risk up front. The entry conditions are a 20,000-historical-ticket bar, a 2,000-a-month floor and an annual commitment made before you know whether it works for you.
Ada's getting-started docs describe what stands in for a trial:
"To help you explore Ada's platform, your Ada representative will provide you with a link to your Agent. This allows you to test key features, interact with the dashboard, and see how your Agent responds to customer inquiries—all before making any long-term commitments."
Forethought's FAQ is direct: "Yes, we offer a Proof of Value (POV) instead of a traditional free trial."
If reading the number before the sales call is the priority, My AskAI publishes its rates: $199 a month on Pro, $499 a month on Scale, and around $0.10 per credit on Scale, with usage-based add-ons metered separately. Our trial runs 30 days with every feature unlocked, unlimited tickets and no card required, so you can test it before you pay.
Conclusion - should I choose Ada CX or Forethought?
TL;DR: Forethought takes five of eleven categories, mostly on reach and tooling. Ada wins on channel breadth and on publishing its security detail openly, and four rows are level. If you run Intercom or HubSpot, the choice is already made for you.
Forethought is the better default for most teams evaluating these two, because it reaches your helpdesk and your documentation wherever they live. Ada takes it when native social channels or a downloadable compliance pack decide the deal.
The eleven categories are modes, ease of setup, training and integrations, answer quality, improving, features, price, customization, lock-in, other AI features and security. Evidence comes from vendor documentation, published pricing pages, dated help-center articles and G2 reviews (Vendr's procurement records are the only observed contract data on either side). Where a vendor publishes no rate, the absence itself is the price evidence, and any claim with no public source carries no score.
Ada CX | Forethought | ||
Modes | 8/10 | 7/10 | Ada win |
Ease of setup | 5/10 | 6/10 | Forethought win |
Training/Integrations | 5/10 | 9/10 | Forethought win |
Answer Quality | 7/10 | 7/10 | Tie |
Improving | 8/10 | 8/10 | Tie |
Features | 7/10 | 8/10 | Forethought win |
Price | 5/10 | 5/10 | Tie |
Customization | 8/10 | 8/10 | Tie |
Lock-in | 6/10 | 7/10 | Forethought win |
Other AI features | 6/10 | 8/10 | Forethought win |
Security | 8/10 | 7/10 | Ada win |
Forethought takes five, Ada two, and four rows are level. Forethought's five cluster in reach and tooling (which is where most of the evaluations I see get decided). Ada's two decide it for a buyer who needs native social or a downloadable compliance pack.
Customization and improving both came out level.
On customization, Ada's Playbooks carry a documented step reference, and its MCP Server stages, tests, approves and reverts behavior changes as a change set, which is what you weigh against Autoflows and the Action Builder. On improving, Ada regression-tests the answering agent at 3,000 simulations a day with coaching applied straight from a failure, and Forethought's bulk simulation covers the Triage classifier.
On answer quality, both bands straddle the field median and neither is audited. On price, both sit within about $1,000 of each other at Vendr's median, and neither publishes a number of its own.
Choose Ada CX if:
- You need WhatsApp, Messenger and Instagram as native channels
- Your stack is contact-center-first, on NICE, Genesys, Amazon Connect or Twilio Flex
- Your security reviewer wants downloadable SOC 3, PCI AOC and pen-test reports before signing
- You want the vendor to stand the agent up for you, since Ada's docs say its own CX team usually configures the AI Agent in advance, adding Knowledge and sometimes the Persona
Choose Forethought if:
- You run Intercom, HubSpot, Front, Gorgias, Freshworks, Help Scout or Kustomer, since Ada does not reach the first three
- You want ticket triage and automated QA from one vendor
- Your knowledge lives in Notion, Confluence, Google Drive or SharePoint
- You have 20,000+ historical tickets to train on
My AskAI installs into Zendesk, Intercom, Freshdesk, Freshchat, Gorgias and HubSpot, we publish our rates on our pricing page, and we go live the same day.
We also have roundups of Ada alternatives and of Forethought alternatives, plus a head-to-head of Intercom Fin against Forethought.
Good luck with the evaluation.
FAQs
Is Forethought still a standalone product after the Zendesk acquisition?
Yes. Zendesk's own help center article describes it as a platform that "can operate independent of the Zendesk platform", and states that "Forethought plans are available for purchase for any customer. Contact Zendesk Sales." That commitment was made at the announcement on March 11, 2026, restated at general availability on June 4, 2026, and it stands in Zendesk's live documentation. Its integrations page still lists 23 helpdesks and CRMs, five months after the deal closed.
Can Forethought fully replace Ada CX, and vice versa?
It depends on your stack, and for most teams Forethought replaces Ada (it names 11 more helpdesks and CRMs than Ada does). Ada keeps native WhatsApp, Messenger and Instagram, plus contact-center voice routing into NICE and Genesys.
Ada matches Forethought on answer quality and beats it on channels. On Intercom, HubSpot or Freshdesk, Ada does not reach your helpdesk at all.
If your helpdesk is one of those three, My AskAI installs into it directly and we train on your existing help center and past tickets.
What can each AI agent be trained on?
Both train on help-center articles and past tickets.
Source | Ada CX | Forethought |
Notion | ❌ | ✅ |
Confluence | ❌ | ✅ |
Google Drive | ❌ | ✅ |
SharePoint / OneDrive | ❌ | ✅ |
Shopify | ❌ | ✅ |
MCP | ✅ (admin and authoring) | ✅ (knowledge connector) |
Both have an MCP server. Check which job each one does.
Forethought's is a knowledge connector on its integrations page. Ada's is for administration and authoring, covering bulk test-case creation and CSV export, while its knowledge-integration docs define its ingestion list and route anything outside it through a custom integration.
How long does setup take and do we need developers?
Forethought quotes 30 to 90 days and wants 20,000+ historical tickets plus at least 2,000 tickets a month. Ada runs 8 to 16 weeks for a full enterprise deployment, or 4 to 6 weeks for a chat-only pilot.
Neither needs your developers for the standard build, because both are implemented by the vendor's own team. Both then pair you with a CSM you will keep depending on for changes, so factor that into the running cost.
Which AI support tool gives the best value per conversation?
Neither publishes enough to answer that: Ada's pricing page carries no price, tier or rate, and Forethought's carries no dollar figure at all. Vendr's procurement data puts Ada's median contract at $73,500 a year across 112 purchases, and Forethought's at $74,483 across 69, on a page dated February 2026.
At Ada's median the effective rate works out around $0.245 per conversation. Forethought's observed Solve band is $0.50 to $2.00 per conversation. Both figures come from Vendr's procurement data, so treat them as a starting point for negotiation.
Free trial AI customer support tools I can test before committing
Ada books a consultation and gives you a rep-provided link to a demo agent. Forethought runs a sales-mediated Proof of Value.
If testing before you buy is the deciding factor, we run a 30-day My AskAI trial with all features unlocked, unlimited tickets and no card required.
Can they handle multilingual support and agent translations?
Both do. Ada handles multiple languages across every channel, with documented language switching and manual translation overrides. Forethought's multilingual sits on its entry Team tier after the 2026 repackaging. Solve chat auto-detects the customer's language, while Solve Email needs your CSM to enable it.
How do Ada CX and Forethought compare on security and compliance?
Ada publishes its compliance pack openly. Its trust center carries downloadable SOC 2 Type II, SOC 3, HIPAA, a PCI AOC, AIUC-1 and 2025 pen-test reports, with no ISO 27001 or ISO 42001 (no sales call needed). Forethought holds SOC 2 Type II and HIPAA, aligns with NIST 800-53, and its Trust Report is in-product and customer-gated.
On ISO 27001, Forethought's security page claims compliance and stops short of certification. The acquisition transferred nothing here either: Forethought is named in none of Zendesk's in-scope product lists for any compliance program, on an article updated July 7, 2026. Forethought's own wording is "We are compliant with ISO 27001 and certified for SOC 2."
