What Is Coming Next for Salesforce’s Infamous Help Portal?
There is perhaps no Salesforce tool that has had more scandals than Salesforce’s Help page, with community feedback (and backlash) shaping everything from how users interact with it to the future of its functionality.
It has now been nearly six months since Salesforce reverted the page to its previous state and restored the Search functionality. Since then, the CRM giant seems to have addressed the feedback, and the upset about the Help page has been soothed. Now armed with additional research and guidance, the team is preparing for the next installment of Help page and agent updates, including promising case-resolution changes and more.
The History of the Help Portal
Ever since Agentforce was launched back in October 2024, Salesforce has proudly been the AI platform’s “customer zero”, using its own technology to test and optimize products before they go to market. One of the main ways it does this is through its Help page.
For two decades, Salesforce’s Help page comprised self-help articles and resources for users to sift through, with a search bar at the top to make it easier for them to find what they were looking for. If they couldn’t find what they were looking for, they could submit a support case.
Once Agentforce was announced, its UI changed, with what Salesforce calls its ‘Help Agent’ embedded into the site, so that people could now ask Agentforce for help. The Help Agent works from the site’s existing, extensive data, so Salesforce’s goal was to make it easier and more efficient for users to get what they needed from the site.
Although people did begin using the Help Agent, like with many Agentforce features, the road to adoption was rocky. In December 2024, my colleague Christine shared her honest thoughts on what the agent could do, and a lot was left to be desired. Six months on, it was revisited, and although progress had been made, there was still a long way to go.
Adoption was increasing, but unfortunately, so were reports of hallucinations, inaccuracies, and the general difficulty in submitting a case or deferring to a human rather than AI.
Salesforce Takes Down Search
Then, on September 29, 2025, Salesforce quietly removed the Search functionality from the Help page, but the fallout was anything but quiet.
Once users caught wind of what happened, a help request ticket popped up on IdeaExchange – Salesforce’s official community request forum, authored by Salesforce MVP Tom Bassett.
“Search has been replaced with Agentforce, which can produce unreliable results and takes longer to find the information you need,” he wrote. “Turning Search back on would enable a UI that can be used to find trusted results and would give the power back to users as to whether they want to search using text or a conversation.”
Bernard Slowey, Salesforce’s SVP of Digital Success, was the one behind the idea and subsequently dealt with the feedback. He explained that although data analysis showed an average of only 1.7% of Help sessions engaging with search, Salesforce had heard the message “loud and clear” that they had made the wrong decision.
Search functionality returned on Friday, November 14, 2025, around seven weeks after it was initially removed.
“I made that call with my team, saying look – let’s take search down and go straight to Agentforce,” Bernard told SF Ben. “And my lesson learned was I made that call too soon.”
What’s Coming Next?
At the second Irish Dreamin’ event yesterday, Bernard provided a sneak peek into the next updates for the infamous Help Page on home soil, once again commending a strong feedback loop and research ahead of these changes.
To date, nearly 3.3M conversations on the site have been handled by Agentforce, with 1.9M deferred to humans. Bernard explained that the case resolution currently sits at just 62% – the goal is to get it to 80% by the end of the year.
So far, we know this is going to be pushed for in two main ways: a more personalized experience and new case resolution metrics.
Help Agent to Go In-App
By the end of the month, Salesforce plans to announce the availability of its Help Agent in-app, meaning that the agent is able to answer questions and provide support to users directly from their orgs, building off of context and user-specific data.
Bernard explained that at present, around 90% of conversations with the agent are unauthenticated, making it difficult to provide targeted support.
“The agent doesn’t know who you are when you ask it a question,” he explained. “It tries to answer, but it doesn’t have a lot of context.”
He said that a more personalized experience, like using ChatGPT when you’re logged in, provides more accurate and positive results as it is able to use its memory and context to formulate answers and solutions from its previous conversations with you – which is exactly what Salesforce is trying to emulate.
He revealed that Salesforce is currently getting ready to launch the Help Agent in-app, replacing the ? icon and functionality in orgs with the agent.
“When you click on that question mark, it will launch the agent,” he said. “Now we [will] know who you are – you don’t need to log in, because the context about you is there.”
This could mean that the days of tab-switching to Salesforce Help will be over very soon.
New Case Resolution Metrics
Bernard was also able to provide insight into how the resolution rates for the agent’s conversations will be measured in the future, especially as there has been some confusion over it in the past.
The Help Agent’s resolution rate is currently measured through:
- Looking at the number of people landing on the Help portal.
- Look at how many people started a conversation with Agentforce.
- Three flows work as a result. One fires when the conversation is started and then dropped off. The “abandonment rate” used to be 26% – now it’s 8%.
- The second one is a hand-off to a human.
- The third is resolution – this is worked out through an additional survey at the end, which prompts the user to answer the question “did we solve the problem?”
However, even with these processes that were brought in the middle of last year, Bernard acknowledged that these are already beginning to feel outdated.
Users can still easily click away from the survey if they don’t want to engage with the agent or the site any longer. Not only that, but users may be feeling frustrated if the agent is either being too specific or not specific enough due to a lack of context – something that they may not necessarily share out of frustration.
“[At the moment] we still have no idea what the skill set of that personal user is,” Bernard explained. “That’s why we’re so big on solving this problem.”
He revealed that at some point in the future, the team would be reevaluating the way it collects user feedback data, with one potential being analyzing a person’s conversation with the agent and basing success on that.
Final Thoughts: Bringing Everything Together
Although we do not have too much information on the concrete details or timelines of these next updates for Salesforce Help, these insights indicate a prominent shift in both Salesforce’s understanding of its user base and its users’ relationship with artificial intelligence. Of course, some devil will inevitably be in the details when dealing with user identity and context across environments such as production versus sandbox orgs.
These updates, at least personally, demonstrate that Salesforce is a lot more clued in to what its community has been asking for, making good use of frequent feedback sessions, testing, and learning from mistakes.
As more information becomes available, we will be sure to update you.
For more Salesforce conversations and industry insights, check out the latest episode of the Picklist podcast below.
