The Smartest New ColleagueEver
Think of general AI as a brilliant new colleague on their first day on the job. Impressively well-read, quick-witted, and articulate. But not quite ready to work yet. That colleague doesn’t know who your most important clients are. They don’t know that billing for project clients works differently than for license clients. They don’t know that the discount your sales director negotiated last month completely changes the renewal conversation.
What that colleague needs is context. This concern is also reflected in research. Salesforce found that a lack of contextual knowledge is among teams’ top concerns regarding AI, right alongside security (Salesforce, State of Service). People intuitively sense that an assistant without context is an assistant you can’t rely on.
Context is where AI really comes into play
Context means that the AI knows what your best people know: which customer the conversation is about. What was agreed upon. What happened last time. Which process applies and who needs to sign off. This is exactly where Agentforce, the platform Salesforce built for AI agents, makes the difference. Many tools retrieve context through separate integrations. An Agentforce agent works directly within the environment where your data already resides. It doesn’t operate alongside your organization. It operates within it. It works from data you can securely connect to, follows your business rules, and acts within the boundaries you define.
Take a lead that comes in through your website. A generic tool might draft a neat, polite response. An agent with context does something entirely different. Even during the preparation phase, it recognizes that this company had already made contact two years ago. It sees which conversation was taking place back then, what proposal was on the table at the time, and why it didn’t go through. It sees which proposal is a good fit now and which colleague handled the contact previously. Where you used to spend half an hour searching through old emails and reports, you now go into the conversation fully prepared. Same kind of technology. Completely different outcome. That’s because context has been added to the mix.
Two types of knowledge, one of which isn’t documented anywhere
Your organization captures knowledge in two ways. The first type is stored in records: customer data, project rules, invoices, and hours. Neatly structured and searchable.
The second type arises from engagement. In the conversation where that discount was agreed upon, or in the thread where your team decided to adjust the scope. In the note that this customer always pays three weeks late. That knowledge explains the records, but it isn’t stored in any field.
A generic model recognizes the first type but overlooks the second. That’s why an answer might sound correct yet still feel out of touch with reality. The record states what was agreed upon. Only the engagement explains why.
The conversation lives in your system of engagement
This shifts the focus to the place where that second type of knowledge is created. Your system of engagement. Most knowledge work doesn’t happen in dashboards. It happens in the channels where your team collaborates. Whether that’s Slack or Microsoft Teams. That’s where decisions are made, questions are asked, and problems first come to light. That’s exactly why an agent belongs right there. If it’s embedded in the conversation itself, you don’t have to leave your work to use it. You ask the question right where it arises. What’s the status of this deal? Has this customer reported this problem before? The agent responds right in the channel, backed by your data. AI is then no longer a separate destination, but part of how work flows.
Here’s an example from our own experience. For a scope proposal, the context was scattered across three places: in an email, in conversation logs in Salesforce, and in a canvas in Slack. We asked our Slackbot, the Agentforce agent that operates as a bot in our Slack channels, to handle it. It combined the structured data from Salesforce with the unstructured notes from Slack and delivered a single, clear scope proposal. Could a generic tool like ChatGPT have done this? Maybe, if we had connected all our apps to it. But then you’d be sharing customer data with tools that aren’t designed for that purpose. Agentforce solves this with a secure middle layer based on the zero-copy principle. Your data stays where it is, yet the agent has all the context they need. That’s exactly where the benefit lies.
What this means for your organization
If generic AI tools have disappointed you so far, the conclusion isn’t that AI doesn’t work. The conclusion is that AI doesn’t work without context. The right question isn’t which model is the smartest. The right question is closer to home. Does the AI know what your best people know? Can it see your customers, your appointments, and your processes? Does it operate within your rules? Answer those questions well, and things will start to shift. AI will then no longer produce generic text, but will perform specific tasks.
On September 9 and 10, the CBEE Remarkable team will be at Data Expo at the Jaarbeurs Utrecht. If you’d like to see what an agent with real context looks like in practice, stop by our booth. We’d love to talk to you about it.
This blog post is a contribution from CBEE Remarkable, an implementation partner specializing in Salesforce and Certinia. CBEE helps organizations in professional services, IT, and Health & Life Sciences digitize and connect their commercial and financial processes, from lead to cash. Find more inspiration at www.cbeeremarkable.nl or visit CBEE Remarkable during Data Expo.
Author: Chris Boonstra
