If you need a Forethought alternative for complex support workflows in 2026, shortlist Sierra, Lorikeet, Decagon and Fin (now part of Salesforce) first, because each publishes a way to run multi-step processes that take actions in your backend systems under deterministic control. Forethought itself is now part of Zendesk, so for Zendesk-first teams the honest default is to stay and test Forethought AI Agents by Zendesk before switching. Agentforce, Maven AGI and Ada round out the list for CRM-native, telephony-heavy and large enterprise teams.

Every vendor fact below was checked on the vendor's own public pages on 30 October 2026 and is linked to its source.

Key takeaways

• Forethought is now Zendesk's. Zendesk announced the deal on 11 March 2026 and completed it on 26 March 2026; the product is now sold as Forethought AI Agents by Zendesk, which Zendesk says work within Zendesk and on any platform.

• "Complex" means actions, rules and proof. A complex workflow verifies the customer, reads and writes to backend systems, applies policy logic, and leaves a record you can audit. Answering from a help centre is not enough: Gartner found only 14% of customer service issues are fully resolved in self-service.

• Several alternatives now publish deterministic controls. Sierra, Lorikeet, Decagon and Fin all describe mixing natural-language instructions with code or controlled system calls for strict steps.

• Published pricing is the exception. Fin ($0.99 per outcome), Agentforce ($2 per conversation) and Lorikeet (from $2,100 a month plus per-resolution rates) publish prices; Forethought, Sierra, Decagon, Maven AGI and Ada quote on request.

Which tools handle complex support workflows?

Lorikeet runs multi-step support workflows that act through your APIs, webhooks and MCP servers, with payments and identity checks executed as deterministic code the agent can invoke but never alter (how it works). Other tools worth comparing:

• Sierra says interactions with systems of record are "deterministic and controlled" (Sierra trust).

• Decagon defines workflows as natural-language Agent Operating Procedures that combine flexibility "with the reliability of code" (Decagon AOPs).

• Fin (now part of Salesforce) uses Procedures with if/else logic, code steps and system connections via data connectors or MCP (Fin Procedures).

What happened to Forethought?

Forethought's homepage now carries the banner "Forethought is now part of Zendesk" (forethought.ai). Forethought's own announcement post says it has "entered into an agreement to be acquired by Zendesk" (Forethought blog), and Zendesk's newsroom confirms the deal closed on 26 March 2026 after being announced on 11 March 2026 (Zendesk newsroom). Zendesk's pricing FAQ describes Forethought AI agents by Zendesk as agents that "can handle more complex workflows across all channels and environments" and work on any platform (Zendesk pricing).

The product line is still listed on Forethought's site: Solve (omnichannel resolution), Triage (ticket classification), Discover (knowledge gaps), QA and an agent copilot, across chat, email, voice, headless and Slack. Its pricing page lists Team, Professional and Enterprise plans, all "Get a Quote" (Forethought pricing). Nothing on these pages says the product is being retired, so do not assume it is. The real question is whether you want your AI roadmap tied to Zendesk's.

How we chose and checked these

We kept platforms that publish a way to resolve multi-step tickets with actions in other systems, not only knowledge answers, and that cover at least chat and email. We compared them on how workflows are written, how strict steps are enforced, how behaviour is tested before launch, published pricing, and helpdesk fit. Every vendor fact was checked on the vendor's own public pages on 30 October 2026, and anything a vendor does not publish is marked "Not published".

Lorikeet publishes this guide and is one of the tools listed. We have tried to describe every product, including our own, using only what each vendor publishes, and we link to the source for each claim.

Forethought alternatives compared

1. Sierra: enterprise depth and payment compliance

Sierra builds agents that connect to systems of record "from order management to CRM" to handle tasks "like processing an insurance claim, returning an order, or originating a mortgage" (Sierra product). In Agent Studio, teams compose step-by-step "journeys", generate them from existing operating procedures, and debug with real-time traces of every decision and tool call. Sierra lists SOC 2, HIPAA, PCI, FedRAMP High, ISO 27001 and ISO 42001, holds PCI DSS Level 1 Service Provider certification, and says interactions with systems of record are deterministic and controlled.

Best for: large consumer brands with complex, transactional workflows, especially ones that take payments in conversation.

Trade-offs: Sierra prices on outcomes but publishes no rate card, and its model leans on its own agent development team, which suits some buyers more than teams that want to edit everything themselves.

2. Lorikeet: complex and regulated workflows with published pricing

Lorikeet is built for support where a ticket means work in another system: checking a loan, changing a payment plan, verifying identity. Workflows act through APIs, webhooks and MCP servers; strict steps run as deterministic code; and you test on replays of historical tickets before go-live, then Coach scores every AI-handled conversation (how it works). It runs on Zendesk, Intercom, Salesforce and other helpdesks, across chat, email, voice, SMS and WhatsApp.

Carmoola, an FCA-regulated UK car finance lender, previously ran automation on a knowledge base and rigid form-driven prompts that resolved about 30% of inbound questions. Its agent resolved 40% end to end from day one and now resolves 60% of all inbound on WhatsApp, chat and email.

Pricing is published: Start $2,100 a month and Scale $5,100 a month, paid annually, plus $0.99 or $0.90 per chat, email or SMS resolution; escalations and unresolved tickets are not charged; 30-day free trial (pricing). The trust page lists SOC 2 Type II, ISO 27001:2022 and HIPAA with BAAs.

Limits: no PCI DSS, so card capture must stay in your own payment flow, and the platform fee makes it poor value for a queue that is mostly simple FAQs.

3. Decagon: natural-language procedures your team owns

Decagon's Agent Operating Procedures let CX teams "define agent behavior in natural language, the same way you train human agents with SOPs", while technical teams version them with Git-based tracking and keep ownership of code (Decagon AOPs). The platform covers chat, voice and email in one layer and includes testing, observability and experimentation (decagon.ai).

Best for: enterprises that want their own team, not a vendor, to iterate on complex workflows quickly.

Trade-offs: no public rate card, so pricing comparison requires a sales process.

4. Fin (now part of Salesforce): Procedures on any helpdesk

Fin, formerly Intercom, became part of Salesforce on 10 September 2026 (Salesforce announcement). For complex work, Fin Procedures let teams write processes in natural language, add if/else conditions and use code "to enforce rules like verifying eligibility, calculating dates, or updating records accurately", connect systems through data connectors or MCP, and run simulations for regression testing (Fin Procedures). Fin works on Intercom and other helpdesks, including Zendesk, at $0.99 per outcome with a 14-day trial (Fin pricing).

Best for: teams wanting transparent per-outcome pricing and a fast trial.

Trade-offs: Intercom's help centre notes that Procedure actions run one by one with no parallel processing. A Procedure ending in a configured handoff counts as a billable outcome.

5. Staying on Forethought AI Agents by Zendesk

For teams whose system of record is Zendesk, staying is a legitimate option. Zendesk includes AI agents in every Suite and Support plan, billed per automated resolution, and describes Forethought's agents as able to handle more complex workflows across all channels (Zendesk AI agents). One contract and native reporting are real advantages. For day-to-day operating detail, see running an AI agent inside Zendesk.

Trade-offs: if your helpdesk is Salesforce, Intercom or another platform, buying from Zendesk means your AI vendor also sells a competing helpdesk. Pricing for Forethought tiers is quote only.

6. Salesforce Agentforce: complex workflows in the CRM

Agentforce suits complex workflows whose data already lives in Salesforce. It publishes consumption pricing via Flex Credits, per-user licensing, or $2 per conversation (Agentforce pricing).

Trade-offs: value depends on Salesforce data and admin skills, and per-conversation pricing is not tied to resolution.

7. Maven AGI: actions across existing helpdesk, CRM and telephony

Maven AGI loads "the right instructions, knowledge, and actions" for each conversation type, integrates with Zendesk, Salesforce and Freshdesk, and runs voice on Twilio, Genesys or Cisco telephony (mavenagi.com). Pricing is not published, and its headline results are vendor-reported.

8. Ada: enterprise SOP automation

Ada offers Playbooks to "automate your complex SOPs with agentic AI" and Ada Computer to give the agent tools to complete tasks, across voice, messaging and email (ada.cx). It suits large, multilingual enterprises; pricing is not published.

What makes a support workflow complex enough to need a new platform?

A workflow is complex when the right answer depends on data the customer cannot see and a rule the agent must apply. In practice that means four capabilities:

1. Actions in backend systems, such as refunds, plan changes or account updates, often chained.

2. Authentication before account data is shared, with strict steps that cannot be talked around.

3. Policy logic that branches on eligibility, jurisdiction or dates.

4. Audit and testing, so you can replay what the agent did and test changes before release.

The CFPB found that chatbots "may be useful for resolving basic inquiries, but their effectiveness wanes as problems become more complex" (CFPB). Gartner's survey found that even for issues customers described as "very simple", only 36% resolved fully in self-service (Gartner). For more on the difference, see agentic AI vs workflow chatbots.

How to choose a Forethought alternative for complex workflows

• Test on your hardest 200 tickets, not your FAQ set, and ask each vendor to show every tool call on a failed case.

• Ask how prompt injection is contained. OWASP says it is "unclear if there are fool-proof methods of prevention", so the question is what a manipulated conversation can reach (OWASP LLM01).

• Map controls to a framework your risk team knows, such as the voluntary NIST AI Risk Management Framework.

• Decide where the helpdesk sits. If you want AI that works on top of the helpdesk you keep, read AI layer vs helpdesk built-in AI.

• Budget for failure modes. Gartner predicts over 40% of agentic AI projects will be canceled by the end of 2027, citing costs, unclear value or inadequate risk controls (Gartner).

What still needs a human?

Complex does not mean fully automatable. Keep people on hardship and vulnerability cases, complaints with regulatory weight, fraud and disputes that need investigation, and any request that touches clinical judgement. Write escalation rules before automation rules: hand over with context when the customer asks for a person, when the agent stops making progress, or when policy requires a human decision. Compliance sign-off on new or changed workflows should stay with your compliance team.

If your hardest workflows match this profile, start a 30-day Lorikeet free trial and replay your own escalated tickets before you commit.