Forethought Solve setup: how do we configure it to resolve common issues end-to-end in chat and hand off to agents when needed?

Most support leaders don’t actually want “a chatbot.” They want Forethought Solve configured so it quietly resolves the top 30–60% of chat volume end-to-end, then hands everything else to agents with full context—without creating a maintenance nightmare.

Below is how I recommend configuring Solve to do exactly that, based on what’s worked across dozens of enterprise deployments where deflection, CSAT, and time-to-resolution are board-level metrics.

---

Quick Answer: The Best Way to Configure Forethought Solve

Quick Answer: The best overall configuration for end-to-end resolution in chat is a policy-bound, Autoflow-driven Solve setup. If your priority is clean triage and routing to agents, a Solve + Triage configuration is often a stronger fit. For teams focused on agent productivity and complex handoffs, consider Solve tightly paired with Assist and Discover.

---

At-a-Glance Comparison

Rank	Option	Best For	Primary Strength	Watch Out For
1	Solve + Autoflows (policy-bound)	High deflection and end-to-end self-service in chat	Resolves common issues directly using actions in your systems	Requires clear policies and initial Autoflow design
2	Solve + Triage	Clean handoffs and routing when AI can’t resolve	Smart classification, prioritization, and agent matching	Less end-to-end automation if Autoflows are minimal
3	Solve + Assist + Discover	Complex operations and continuous improvement	Agents and AI co-resolve while Discover closes knowledge gaps	Needs coordination across Ops, Knowledge, and Training teams

---

Comparison Criteria

We evaluated each setup against three practical criteria:

• Resolution Rate & Deflection:
  How much of your chat volume is actually solved end-to-end by AI (not just “answered”) before an agent is needed.

• Operational Overhead & Maintainability:
  How easy it is to deploy, govern, and update over time without building a new decision tree for every scenario.

• Handoff Quality & Agent Efficiency:
  When Solve does hand off, how much context the agent gets and how much faster they can resolve the issue.

---

Detailed Breakdown

1. Solve + Autoflows (Best overall for high deflection and end-to-end resolution)

Solve + Autoflows ranks as the top configuration because it’s the most direct path to resolving common issues end-to-end in chat while staying within your business policies.

Instead of just answering questions, Solve can reason, decide, and take action—executing Autoflows across your tech stack to actually fix the issue (e.g., reset passwords, update orders, process cancellations) without an agent involved.

What it does well:

• End-to-end resolution in chat:

  • Design Autoflows around your top use cases (e.g., refunds within policy, subscription changes, shipping status).
  • Solve uses your past tickets and help center content to understand intent, then triggers the right Autoflow.
  • Actions are taken in your systems via integrations and API connectors (e.g., update a record in Salesforce, issue a refund via your billing provider) so the customer doesn’t need to be transferred.

• Policy-bound actions (you stay in control):

  • Every Autoflow is governed by your business rules—eligibility checks, limits, approval thresholds.
  • You can restrict certain actions to specific conditions (e.g., “only offer partial refunds for orders under $200 and within 30 days”).
  • Role-based access and audit-ready logs ensure you can see who changed what and when, which matters for finance, legal, and security teams.

Tradeoffs & Limitations:

• Upfront design work (but low ongoing upkeep):
  • You will need to invest upfront in mapping your top 10–20 chat scenarios into Autoflows.
  • The good news: you’re not scripting every path; Solve uses your data and reasoning to handle variation in language and edge cases.
  • You’ll want a clear owner in Support Ops or CX to coordinate policies with stakeholders like Finance, RevOps, and Product.

Decision Trigger:
Choose Solve + Autoflows if you want maximum deflection and resolution in chat, and you’re ready to codify key policies into Autoflows so the AI can take real actions, not just respond.

---

2. Solve + Triage (Best for clean routing and controlled handoffs)

Solve + Triage is the strongest fit when your priority is getting the right issues to the right agents while still resolving a healthy portion of chat traffic up front.

Solve handles common issues in chat; when it can’t resolve the request under your policies, Triage classifies and routes the ticket to the best queue, priority, or specialist team.

What it does well:

• Smart classification and routing:

  • Triage auto-tags and prioritizes conversations based on historical ticket patterns, sentiment, and context.
  • You can use those tags to route high-value or high-risk issues (e.g., billing disputes, VIP accounts) directly to senior agents or specialist pods.
  • This reduces time-to-resolution because agents see the right tickets faster, with less manual triage.

• Controlled escalation from chat:

  • When Solve determines a case should be escalated (due to complexity or policy limits), it passes structured data into the helpdesk.
  • Triage then classifies the issue, attaches context, and routes it according to your rules, so there’s no “dead air” between chat and the queue.
  • Combined, Solve + Triage reduce “back-and-forth” and handle a larger portion of your volume efficiently.

Tradeoffs & Limitations:

• Less automation without Autoflows:
  • If you don’t pair this with well-designed Autoflows, you’ll get great routing but less end-to-end automation.
  • The result: better agent efficiency and lower response times, but lower deflection compared to a fully agentic setup.

Decision Trigger:
Choose Solve + Triage if you want strong handoffs, clean queues, and better agent utilization, and you’re okay with AI resolving fewer cases directly while you build toward more Autoflows over time.

---

3. Solve + Assist + Discover (Best for complex operations and continuous improvement)

Solve + Assist + Discover stands out for teams that treat automation as a continuous program, not a one-time launch.

Solve resolves what it can in chat. When handoff is needed, Assist supports agents inside the helpdesk, and Discover turns those interactions into recommendations to improve both AI flows and your knowledge base.

What it does well:

• Agent and AI co-resolve complex issues:

  • Assist gives agents in-tool AI support: ticket summaries, suggested replies, answer drafting based on your help center and past tickets.
  • This compresses handling time and keeps responses on-brand, even on complex or long-running issues.
  • You get not just deflection via Solve, but faster resolution when humans are involved.

• Continuous improvement with Discover:

  • Discover analyzes interactions (resolved by both AI and humans) to find knowledge gaps and workflow opportunities.
  • It surfaces “questions the AI couldn’t confidently answer” and areas where agents are improvising workarounds.
  • You can then generate new articles, update policies, or design new Autoflows—which improves Solve’s resolution rate over time.

Tradeoffs & Limitations:

• Program-level coordination required:
  • This configuration delivers the most strategic benefit, but it does require alignment across Support Ops, Knowledge Management, and Training.
  • You’ll get more value if you treat Discover’s insights as a backlog for monthly improvement cycles, not just a dashboard to glance at.

Decision Trigger:
Choose Solve + Assist + Discover if you want a long-term, compounding lift in resolution rate and agent efficiency, and you’re prepared to run an ongoing AI improvement program.

---

How to Configure Solve to Resolve Common Issues End-to-End in Chat

Now let’s move from “which setup” to the actual configuration steps. This is the practical blueprint I use with teams implementing Forethought Solve for end-to-end chat resolution and smart handoffs.

Step 1: Identify the Right Use Cases for End-to-End Resolution

Start by targeting issues that are:

• High volume
• Low to medium complexity
• Governed by clear policies
• Safe to automate with appropriate checks

Examples:

• Order status and tracking
• Password resets and account lockouts
• Refunds and returns within policy
• Subscription changes (upgrade, downgrade, cancel)
• Profile/contact information updates
• Basic billing questions (invoices, past charges, payment methods)

Use your past ticket data (which Forethought already learns from) plus Discover insights to rank these use cases by volume and impact. Aim for your first wave of Autoflows to cover issues that can realistically drive 30–40% deflection.

Step 2: Connect Solve to Your Existing Stack

Forethought is designed to work within your existing workflows—no need to rip and replace.

Typical connections:

• Helpdesk / CRM: Zendesk, Salesforce, Freshdesk, Intercom (for tickets, user data, and context).
• Order/Subscription Systems: Shopify, Chargebee, Stripe, Recharge, or custom backends via API connectors.
• Identity Systems: SSO/IdP, authentication services for secure account actions.
• Knowledge Base: Your help center content, product docs, and internal KB.

These integrations allow Autoflows to take real actions, such as:

• Fetching order details and updating shipping addresses
• Issuing partial or full refunds within policy
• Changing subscription plans or renewing services
• Creating or updating tickets with structured fields

Step 3: Configure Solve’s Policy-Bound Autoflows

For each targeted use case, define a clear policy and then translate that into an Autoflow.

Example: Refund within policy

• Policy rules:

  • Orders within 30 days and under $200 are eligible for automatic refund.
  • Anything else escalates to an agent.

• Autoflow design:

  • Solve verifies identity (order number, email, or secure login).
  • Checks order details via your ecommerce/billing integration.
  • If order meets policy rules: execute refund and send confirmation in chat.
  • If not: branch to escalation flow and hand off to an agent with full context.

You repeat this structure for each use case, always asking:

“What decisions are safe to automate under our policies, and what must go to an agent?”

This is where Forethought’s “fully agentic” approach is critical—Solve is not just matching FAQ answers, it’s reasoning with your policies and acting in your systems via Autoflows.

Step 4: Configure Handoff Rules and Escalation Paths

End-to-end resolution in chat does not mean “never escalate.” It means you escalate intelligently.

Define when Solve should hand off:

• Confidence thresholds:

  • If Solve’s confidence in an answer or action drops below an agreed threshold (e.g., 80%), it should escalate.
  • This is backed by Hallucination Mitigation, where Solve verifies facts before responding, reducing the risk of “confidently wrong” answers.

• Policy exceptions and risk scenarios:

  • High-value accounts, high-dollar refunds, legal or compliance-related questions can be flagged for mandatory escalation.
  • Solve can inform the customer that it’s connecting them to a specialist, setting expectations clearly.

• Channel preferences and business hours:

  • Outside agent hours, Solve can provide the best possible answer and create a ticket for follow-up.
  • During hours, it can transition the chat to a live agent with context, or convert to a ticket routed via Triage.

When handoff happens, ensure:

• Context is passed automatically:

  • Conversation history, user identity, detected intent, and any actions already taken by Solve.
  • This cuts repeat explanation and reduces handle time.

• Triage rules are applied (if enabled):

  • Tags and priorities determine which queue or team receives the request.
  • VIP or high-risk cases are surfaced quickly to protect CSAT and revenue.

Step 5: Set Guardrails, Governance, and Trust Controls

For enterprise teams, the right governance makes or breaks rollout. Configure:

• Role-based access and permissions:

  • Limit who can publish new Autoflows, adjust policies, or change thresholds.
  • Create a clear review process for high-impact flows (billing, legal, healthcare).

• Compliance and security posture:

  • Forethought adheres to SOC 2 Type II, HIPAA, GDPR, CCPA, and NIST Cybersecurity Framework.
  • Ensure your legal and security teams review data-handling settings, encryption, and audit logs.

• Hallucination mitigation settings:

  • Keep “fact verification before responding” enabled, especially for regulated or sensitive domains.
  • Decide which data sources are “authoritative” for answers so the AI doesn’t improvise.

These controls give your stakeholders confidence while you expand automation coverage.

Step 6: Measure, Iterate, and Expand Coverage

Once Solve is live with your initial Autoflows, you need a feedback loop. Focus on:

• Key metrics:

  • Deflection rate (AI-resolved vs. agent-handled)
  • First response time (chat)
  • Time-to-resolution (AI + agent combined)
  • CSAT across AI-only, agent-only, and blended interactions
  • Resolution rate tied to specific Autoflows

• Discover-led improvement:

  • Use Discover to identify:
    • Common questions Solve can’t yet resolve.
    • Articles that need updates or don’t exist.
    • Patterns where agents follow the same steps repeatedly (candidates for new Autoflows).
  • Turn those insights into monthly improvements to content and flows.

• Progressive rollout:

  • Start with non-critical use cases and ramp up as metrics prove out.
  • Run a Proof of Value (POV) period where you compare “with Solve” vs. “control” to quantify ROI.
  • Expand Autoflows into more complex, higher-impact workflows as confidence grows.

Over time, this is how teams reach up to 98% resolution rates on targeted intents and see 15x ROI from a combination of deflection, faster handling, and reduced operational debt.

---

How Forethought Solve Handles Multi-Channel and Voice Handoffs

While this article focuses on chat, the same configuration patterns apply across channels:

• Chat and web messaging:

  • Solve resolves issues in real time.
  • Escalates to agents via chat or ticket creation with Triage routing.

• Email and ticket-based interactions:

  • Solve can automatically respond to eligible tickets using the same policies and Autoflows.
  • When uncertain, it drafts responses for agents via Assist, speeding up resolution.

• Voice and phone support:

  • Solve can power voice-enabled flows or follow up via SMS/email after the call.
  • Tickets originating in voice channels are enriched and routed via Triage, then assisted by Assist for quicker handling.

The goal is a consistent, on-brand experience across channels, with the same policies and Autoflows underlying both AI and human responses.

---

Final Verdict

To configure Forethought Solve so it resolves common issues end-to-end in chat and hands off to agents when needed, you should:

1. Design around your highest-volume, policy-clear use cases for end-to-end resolution.
2. Connect Solve to your existing systems so Autoflows can perform real actions, not just send answers.
3. Build policy-bound Autoflows that automate safe decisions and escalate the rest.
4. Configure smart handoffs with Triage and Assist so agents get full context and move faster.
5. Enforce governance, security, and hallucination mitigation for enterprise-ready control.
6. Use Discover and metrics to iterate, expanding automation based on real interaction data.

That’s how you move from “a chatbot that answers FAQs” to an agentic AI system that resolves more, escalates less, and keeps your support team focused on the work that truly needs a human.

---

Next Step

Get Started