Gorgias has been teasing AI Agent 3.0 for several weeks, with much of the promotion centered on Gaia, its new coaching agent. We watched the live event and picked out the latest developments for you, with commentary on what’s competitive and what isn’t.
Why Gorgias rebuilt its AI Agent
The Gorgias team shared that with AI Agent 2.0, it was difficult to have confidence in responses at scale. Unstructured knowledge contributed to inconsistent responses, and a rigid, linear workflow approach could cause difficulty in resolving inquiries that didn’t precisely follow that path.
AI Agent 3.0 represents a shift in focus from generating replies to driving resolutions, with a central reasoning layer and intent-based architecture to drive improved performance.
What are the actual changes in AI Agent 3.0?
The name of the game is Skills and Actions. Gorgias AI Agent 2.0 focused on creating “guidance” that the agent could search through to figure out how to resolve a ticket; AI Agent 3.0 shifts to using Skills that are supplemented by guidance. Each Skill is associated with a specific intent (e.g. order question, refund / cancellation, etc.) along with a prompt that defines how that type of question should be handled.
Guidance has been repurposed as general reference knowledge that can still be accessed by the AI agent.
Gorgias is also putting a big focus on Actions, which are connected via apps you install and can be things like order edits in Shopify, subscription skips or changes, and more.
What about Gaia?
The other big focus is on Gaia, an AI agent that’s designed for use by your internal team, with access to your Gorgias workspace and AI Agent configuration.
You can use Gaia to analyze your tickets, set up Skills and Actions, and generally review and modify the agent’s configuration through a conversational interface.
TikTok Shop integration
Gorgias also announced a native TikTok Shop integration, but their CEO Romain shared in chat that while it can be connected to Gorgias, AI Agent can’t interact with it yet.
Reliance on prebuilt integrations becomes a limitation when a required workflow is not yet supported, especially if there’s no committed timeline. Valiopt delivers custom integrations for our customers when they need them. Most of our integrations are turned around from request to delivery within a week.
What questions are Gorgias users asking?
Throughout the webinar, we took note of questions that Gorgias users were asking via chat. We’ve highlighted some of them here.
We saw a lot of these types of messages in the chat from a number of different users. It looks like there might still be some reliability issues here.
This brings up a great point around data integration with international carriers. Often this tracking info isn’t available in Shopify, and the default apps that Gorgias supports for tracking integration can be limited.
Valiopt solves this for our customers by setting up custom order retrieval integrations that can augment Shopify data with detailed tracking info from a number of third-party sources.
This brings up a great point around differentiated handling for inquiries, aside from intent-based prompts. The Gorgias AI agent is limited here; Valiopt can help you set up unique handling guidance based on a number of custom factors, including customer region.
Gorgias doesn’t currently expose this. They do have a playground that you can ask questions in, but it’s challenging to preview a response to a real ticket. Valiopt supports native preview generation on live tickets, without altering production data or sending responses, so you can test changes against historical inquiries.
Based on our testing, it’s not possible to create custom intents, you’re limited to what Gorgias provides (things like order updates, refunds, cancellations, etc.). This is tricky if you do have common request categories that are more specific.
Valiopt has supported this type of custom intent configuration from day one. We set up unique handling guidance for every type of request you receive, enabling configuration with high specificity.
This is an interesting point that we’ve seen repeated a number of times: CX managers often find setting up these AI agents to be challenging, and for good reason. These are complex technical systems, and it’s not always clear how you can get the best performance out of them.
Gorgias doesn’t offer hands-on support for most of their customers. Gaia is an attempt to solve for this using AI, but as we all know sometimes you still need human guidance. Gorgias has third-party partners that help with some of this (at a cost), but those are folks outside of the company.
Valiopt solves for this by supplementing technology with a human touch that’s always available. We configure automations for you, provide guidance on what works and what doesn’t, and offer ongoing support whenever it’s needed.
This is a great point as it relates to built-in, “default” actions provided by Gorgias. They work for the simplest Shopify setup, but you can’t get much customization out of it.
Valiopt sets up custom integrations for each of our customers that take their unique business needs into account. Multiple inventory locations with custom rules? Not a problem.
Similar to above, there are many integration limitations that are a result of having a one-size-fits-all “Action” for fetching data. This might work for many customers, but it won’t work for all. Valiopt can set up custom integrations that work the way your business wants them to.
Now this is a big one; having the ability to edit an order in Shopify is only one part of the battle; most ecom businesses have a third-party warehouse or fulfillment system that also needs to be updated. We wrote about this in more depth in a separate blog post, but in short: using default actions doesn’t work when you have an additional system that needs to be updated. You need a composed action that can perform the update in all of your systems, at once, based on data from each of these systems. Valiopt can deliver this via a custom agent action.
This is an issue we see a lot: in this day and age, everyone has multiple email addresses, and the one you’re reaching out from doesn’t always match the one that was used to place the order. Authentication is tricky here: you don’t want to give out information to someone who isn’t necessarily the customer.
Valiopt can set up custom authentication flows for cases like these, allowing the data to be retrieved securely, without undue inconvenience to the customer.
This is currently not supported with Gorgias AI Agent, but definitely a reasonable ask. With Valiopt, you can configure custom rules like this, in addition to setting up a customized chatbot embed on your shop that’s only shown outside of business hours.
Many e-commerce stores have a large number of SKUs, and querying these directly from the source isn’t always performant. While the Gorgias implementation here may be limited, Valiopt has an extensive custom product indexing pipeline. We ingest product data as it’s changed and created, resulting in a performant and customizable product search experience.
Definitely a fair question, while Shopify might be the top growing platform, brands operating outside Shopify still need comparable integration depth. This brings us back to custom integration support: having a partner like Valiopt that can deliver custom integrations frees you from waiting on a large company to eventually deliver what you’ve been waiting for.
How does the new AI Agent 3.0 compare to other top products in the market?
There’s no doubt these new features improve the functionality of the Gorgias AI Agent, but how do they compare to what else is available on the market, and in particular the type of functionality you can get with more customized platforms like Valiopt.
The new skill / intent-based request classification is a big one, this will no doubt improve performance considerably, but it’s hardly industry leading or novel. Valiopt has been built on this type of architecture from the beginning, but with the added benefit of customizable intents and sub-intent classifications that can improve adherence to complex SOPs.
Actions and integrations are another big focus. It’s great that Gorgias is launching so many built-in actions and there’s no doubt that they’ll be useful for a lot of stores. Where it gets trickier is if there’s any need to customize these integrations or compose them. Valiopt can help set up custom integrations that perform exactly as your business needs, rather than granting broad permission to modify Shopify data.
Gaia is a big value-add for anyone looking for a conversational interface to their Gorgias data. The ability to ask questions about recent ticket volume and get assistance with configuration is huge. When it comes to AI agent setup though, you can end up in situations where an AI training an AI leads to poor performance, and you really want to talk to a human to figure out how to set something up. Valiopt provides a dedicated human contact for all of our implementations.
Want help setting up a more customized, performant AI automation system for your CX operation? Get in touch.