Melty is the Learning Agent that MeltingSpot now runs inside its own platform. It guides the teams who use MeltingSpot, in conversation, at the moment they need help, without sending them to a help center. We built it because we build Learning Agents for others: the most honest way to show what one does is to rely on it ourselves, every day.
What is Melty, and why is MeltingSpot launching it now?
Melty is MeltingSpot's own Learning Agent, embedded in the MeltingSpot product to guide the people who configure and run it. When a Head of Customer Success, an L&D manager or a change lead sets up their own Learning Agent, Melty sits next to them in the interface. It answers questions, points out the next useful step and explains features in context, as a colleague would.
The launch is a deliberate choice of consistency. MeltingSpot asks software vendors and large organizations to trust a Learning Agent with their users. Running the same technology inside our own product means we live with its strengths and its limits daily. Every friction our users meet in MeltingSpot becomes a signal we see first, and a lesson we fold back into the Learning Agent our customers deploy.
Melty also gives a face to an idea that is easier to show than to describe. The name, avatar and personality are ours. When a customer deploys the Learning Agent in their own software, they choose their own name, avatar, tone and colors, so their users see their brand, not ours. Melty is simply what the MeltingSpot Learning Agent looks like when MeltingSpot is the vendor.
Illustrative scenario: Melty inside MeltingSpot
Melty: Hi Claire. You've imported your onboarding guide, but your Learning Agent doesn't know where to intervene yet. Want me to show you how to define a first coaching zone?
Claire: Yes. I want it to help users on our integrations page.
Melty: Open that page with the extension active, then click the area where users get stuck. I'll suggest content from your guide that matches it, and you can adjust the tone before publishing.
This exchange is a fictional example written to show the behavior, not a transcript of a real customer. What matters is the pattern: Melty notices where the admin is, proposes the next step, and lets the person decide. The same pattern drives every MeltingSpot use case described below.
How does a Learning Agent like Melty behave inside a product?
A Learning Agent is proactive, conversational and contextual: it detects friction and starts the conversation, instead of waiting for a question. That is the core difference with a support chatbot, which only helps users who already know what to ask. For a deeper comparison, see our guide to AI agents vs in-app chatbots.
Detection comes first. The MeltingSpot Learning Agent reads usage signals on the pages and fields where it has been told to watch: a user hesitating, skipping a step, entering data in the wrong format or never opening a feature. Those signals tell it who needs help, on what, and when, so the intervention lands at the moment of need.
The format adapts to the situation. MeltingSpot combines four intervention modes: a conversation, a contextual tooltip on a field or button, a video or document section, and micro-learnings organized into learning paths. A quick question gets a two-line answer. A complex process gets the exact video clip. A skill that takes time gets a short path the user completes inside the tool.
Why does conversation reduce cognitive load better than tours and courses?
Conversation reduces cognitive load because it delivers only what the user needs for the task in front of them, in their own words. John Sweller's cognitive load theory, first published in 1988, holds that working memory is limited. Every extra screen, tab or generic explanation competes with the task itself, and learning suffers when that budget is exceeded.
Software products keep getting richer. Each release adds features, workflows and settings that users must absorb on top of their real job. Product tours fire the same sequence for everyone, regardless of what the user is trying to do. Long documentation asks people to leave their work, search, read and come back. Both add load at the worst moment.
Timing matters as much as format. Hermann Ebbinghaus showed in 1885 that memory decays quickly without reinforcement, which is why a course taken weeks before using a tool rarely survives contact with the real task. Our article on the forgetting curve in corporate training details the mechanism and what it costs training budgets.
Time to learn is also scarce. Bersin by Deloitte has noted that employees can dedicate only about 1% of their workweek to formal learning. Help therefore has to come to the work, a principle known as performance support. Learning has always happened through language, which is why we see conversation, not the default e-learning module, as the natural format for learning software.
Ready to boost adoption?
Give your users a Learning Agent that knows your software
Train and support every user in the flow of work. No credit card, free until you go live.
How does the Learning Agent work as a Digital Customer Success Manager?
As a Digital Customer Success Manager, the Learning Agent lives inside a SaaS product and guides every customer to activation, feature adoption and renewal. It scales what the best CSM does for a handful of accounts to every user, including the silent majority who never open a ticket or attend an onboarding call. Melty itself is this use case applied to MeltingSpot.
Who is it for?
This use case serves software vendors, typically mid-market B2B SaaS companies whose customer base grows faster than their Customer Success team. The people concerned are the Head of Customer Success, Customer Success Managers, onboarding and support leads, customer education teams, and on the product side the VP Product and Product Managers who need new features to be discovered and used.
The problems are familiar. Users sign up, explore, fail to find the value and leave quietly. CSMs spend their days answering the same "how do I" questions. Features ship to silence because nobody guides users toward them. The Digital Customer Success Manager addresses these patterns directly inside the product, where they happen.
What does it look like in practice?
Illustrative scenario: a trial user in a project management SaaS
Day 3 of a trial. Marc has created a project but has not invited anyone, the step that usually makes the product useful for a team. The Learning Agent, branded with the vendor's name and avatar, opens a conversation.
Learning Agent: Hi Marc. Projects become much more useful once your team is in. Want me to show you how to invite two colleagues in under a minute?
Marc: Can they see everything if I invite them?
Learning Agent: Only this project, unless you change it. You can set them as viewers or editors. I'll highlight the button.
In this fictional scenario, three things happen that a product tour cannot do. The agent picks the moment from behavior, answers a follow-up question about permissions in context, and leaves the decision to the user. The CS team gets the activation signal without having scheduled a call. Our article on the AI agent for customer success explores this model in depth.
How does the Learning Agent work as a Digital Corporate Trainer?
As a Digital Corporate Trainer, the Learning Agent trains employees on their business software directly inside the tools they use, at the moment they need it. It turns each application into a hands-on training environment, from a new joiner's first week to advanced features, and measures real proficiency rather than course completion.
Who is it for?
This use case serves corporate learning and development teams in mid-size and large organizations whose employees work in CRMs, ERPs, HRIS and collaboration tools every day. The job titles concerned are the Chief Learning Officer, the Enterprise Training Director, the Head of Organizational Development, L&D Managers and Directors of L&D, as well as sales enablement leads who train reps inside the CRM.
The trigger is usually a gap between completion and performance. Courses are finished, completion rates look fine, yet employees return to their desks and work the old way. Tools also evolve every quarter, so LMS content goes stale before anyone opens it. The Digital Corporate Trainer brings training into the workflow, where it can actually be applied. Our comparison of learning in the flow of work vs LMS explains how both coexist.
What does it look like in practice?
Illustrative scenario: a new sales rep in the CRM
Second week in the job. Sofia is creating her first opportunity in Salesforce and moves straight to the amount field without linking an account, which the company's process requires.
Learning Agent: Before the amount, link the account at the top of the form. Without it, the approval flow won't start. Want me to highlight it?
Sofia: Yes. And what stage should I pick after a first call?
Learning Agent: Use Discovery. Here is the two-minute video from your onboarding path that explains each stage.
In this illustration, the employee learns the process by doing it, and the reinforcement arrives during the real task instead of in a classroom two weeks earlier. The same logic applies to AI tools: the Learning Agent can coach employees inside the interfaces where features like Copilot live, as discussed in our guide to Microsoft Copilot adoption.
How does the Learning Agent work as a Digital Change Manager?
As a Digital Change Manager, the Learning Agent makes software rollouts and AI transformations stick by coaching users in real time after go-live, when most change programs stall. It stays in the tool once the formal training ends, catching errors and regressions, and onboarding new arrivals to the same standard as on day one.
Who is it for?
This use case serves large organizations deploying CRM, ERP, HRIS or AI tools across many teams, entities and countries. The people concerned are the Digital Transformation Officer, Heads of Change Management and change managers, transformation program leads, and the Head of Organizational Development, often working alongside the Chief Learning Officer on the training side of the rollout.
The trigger is the go-live gap. Sessions were attended, licenses are paid, yet months later employees still use workarounds and the support desk is flooded. Our analysis of why ERP implementations fail shows how often adoption, not technology, is the weak point. The Digital Change Manager keeps coaching in the tool after go-live.
AI rollouts add a regulatory dimension. Article 4 of the EU AI Act, applicable since February 2025, requires organizations deploying AI systems to ensure a sufficient level of AI literacy among the staff who use them. In-context coaching, with engagement tracking, gives change teams a practical way to train people on AI features where they use them.
What does it look like in practice?
Illustrative scenario: three months after an ERP go-live
The project team has moved on. Karim, a buyer, is entering a purchase order and types the vendor code in the format of the old system. Before he submits, the Learning Agent steps in.
Learning Agent: That vendor code format comes from the previous system. The new one uses a different structure. Here is how to find the right code from the vendor list.
Karim: Will my old purchase orders still work?
Learning Agent: Yes, they were migrated. Only new entries need the new format. I'll remind you if it happens again this week.
In this fictional example, the error is prevented rather than corrected by the support team, and the change team sees which processes still generate friction across entities. That visibility helps defend the transformation in front of a steering committee. Our digital change management guide covers the wider program around this kind of coaching.
Which MeltingSpot use case fits your organization?
The right use case depends on who your users are: customers in your product, employees learning their tools, or teams going through a rollout. The Learning Agent is the same technology in all three cases. What changes is the audience, the trigger and the indicator that proves it works. The table below summarizes the three options side by side.
| Use case | Company type | Job titles | Problem solved | Example trigger | KPI impacted | Learn more |
|---|---|---|---|---|---|---|
| Digital Customer Success Manager | SaaS vendors, mostly mid-market B2B | Head of Customer Success, CSM, onboarding and support leads, customer education, VP Product, Product Manager | Users who never reach value, repetitive questions, features nobody discovers | A trial user stalls before a key activation step | Activation, time-to-value, feature adoption, support tickets, churn | Digital CS Manager |
| Digital Corporate Trainer | Mid-size and large organizations training employees on business software | Chief Learning Officer, Enterprise Training Director, Head of Organizational Development, L&D Manager, sales enablement | Training completed but not applied, content outdated by tool changes | A new hire skips a required step in the CRM | Time-to-proficiency, advanced feature adoption, training cost per employee | Digital Corporate Trainer |
| Digital Change Manager | Large organizations deploying CRM, ERP, HRIS or AI tools | Digital Transformation Officer, Head of Change, change managers, transformation program leads | Adoption that stalls after go-live, workarounds, flooded support | A user enters data in the old system's format | Adoption curve, error rate on critical flows, AI feature usage, ticket volume | Digital Change Manager |
FAQ
Is Melty a chatbot?
No. A chatbot waits for a question and answers it. Melty, like every MeltingSpot Learning Agent, detects friction from usage signals and starts the conversation itself when it can help. It combines conversation with tooltips, videos and micro-learnings, and it aims to make users autonomous rather than to deflect them.
Can we give our own Learning Agent a different name than Melty?
Yes. Melty is the name of the Learning Agent inside MeltingSpot's own product. When you deploy the Learning Agent in your software or your business tools, you choose its name, avatar, personality, tone and colors. Your users see your brand and your context, and MeltingSpot stays in the background as the engine.
Does a Learning Agent replace Customer Success, L&D or change teams?
No. It takes over the repetitive part of the work: recurring explanations, reminders, first-level questions and the basics of each process. That frees CSMs for renewals and expansion, L&D teams for program design, and change teams for stakeholder management, the work where human judgment makes the difference.
How is a Learning Agent deployed, and does it need developers?
No developer is required. The Learning Agent is deployed through a JavaScript snippet in a web product or a browser extension for enterprise software. Admins name the agent, point at the pages and fields where it should intervene, and import existing content such as documentation, PDFs or videos, from which AI generates coaching content. It works in any web-based tool, from a SaaS product to Salesforce, SAP or Workday, and one agent can follow a user across them.
Conclusion
Melty is MeltingSpot's own Learning Agent, launched inside our own product so that we rely daily on what we offer our customers. It guides users in conversation, at the moment of need, to reduce the cognitive load of ever-richer software. The same Learning Agent serves three use cases: Digital Customer Success Manager for SaaS vendors, Digital Corporate Trainer for L&D teams, and Digital Change Manager for transformation programs.

