Salesforce Consultant

5-ian-cosgrove

Ian Cosgrove

for-everyone

For Everyone

For as long as I can remember, our role as [Salesforce Admins](/blog/tips-for-becoming-a-salesforce-admin?utm_source=cloudanswers_website&utm_medium=blog&utm_campaign=agentforce_vibes_an_admins_perspective) has been defined by the no-code/low-code philosophy. We’re the builders of our Salesforce orgs, the ones who leverage clicks-not-code tools like Process Builder, Workflow Rules, and, of course, [Flows](/blog/flows-vs-workflows-vs-process-builder-process-automation-in-salesforce?utm_source=cloudanswers_website&utm_medium=blog&utm_campaign=agentforce_vibes_an_admins_perspective) to automate business processes and customize user experiences. Our superpower? Understanding business needs and translating them into efficient, declarative solutions right within the Salesforce platform.

But what if the boundaries of no-code started to get a little blurry? What if we could not only understand complex [Apex code](https://developer.salesforce.com/docs/atlas.en-us.apexcode.meta/apexcode/apex_dev_guide.htm) but also generate it, debug it, and even deploy it, all without ever having to learn a coding language? This is exactly where [Agentforce Vibes](https://www.salesforce.com/agentforce/) steps in, a game-changer for us, Salesforce Admins.

The big question I’ve been asking myself, and what we’re exploring today, is, can Agentforce truly empower us Admins to become viable coders, or at least highly proficient code-adjacent professionals? Let’s dive in and see what I’ve found.

## The Shift from Traditional Admin to Admin+

The Salesforce ecosystem is always evolving, and with it, the demands on us admins. As our businesses grow more complex, so do our Salesforce implementations. While declarative tools are incredibly powerful, there are always those scenarios where Apex code just becomes necessary to achieve specific, nuanced requirements.

This is where I see the concept of Admin+ really taking shape. An Admin+ isn’t meant to replace a Developer; instead, we’re Admins augmented by AI. Agentforce Vibes fits perfectly into this vision, acting as a skill amplifier for us. It’s not about AI replacing our human expertise, but about enhancing it, allowing us to tackle challenges that were previously out of our reach, making us even more valuable to our organizations.

### Setting Up Agentforce Vibes

Getting started is actually pretty straightforward. Agentforce Vibes can be used either through a dedicated Integrated Development Environment (IDE) or as an environment in Salesforce itself. If you choose to go with Agentforce Vibes in the web browser, it will look like this, allowing you to vibe code right in your org (preferably a sandbox). Either way you set it up, you just have to click the AgentForce Vibes icon to get to vibe coding.

![](/uploads/wordpress/2026/02/image-6-1024x646.png)

### Explaining Code

One of Agentforce’s most valuable features, in my opinion, is its ability to break down complex Apex code into plain, understandable English. As a consultant, there are many hands constantly pushing code into my clients’ orgs. With Agentforce Vibes, I can simply specify the Apex Class or Trigger and ask it what it does, allowing me to get a much clearer picture of how everything works together. This is not only helpful for those who are not code inclined, but for the developers as well. Because instead of spending their time explaining what everything does, it speeds up knowledge transfer, allowing them to get back to coding and the Admins to understand everything quickly.

### My Real-World Example: Bridging the Flow-Code Gap

Let’s walk through a recent project that perfectly illustrates Agentforce’s capabilities. I recently had a client request that seemed straightforward at first; they wanted a custom Lead creation process. I built a screen Flow to handle the user interface and initial logic, but I kept hitting a wall. The core issue was trying to bypass the default Lead conversion logic and existing validation rules that the Flow simply can’t do on its own. I thought I was staring down the barrel of a developer request, thinking that it was going to be a bigger project than it turned out to be.

Then, I remembered Agentforce Vibes. I opened VS Code, pulled my existing Flow into the environment, and then turned to Agentforce. I instructed it to create an Apex Class that could plug into the Flow and create records, using the Flow variables. Agentforce didn’t just generate code, it became my coding partner. It created two Apex Classes to handle the two possible outcomes and inserted them into the Flow by updating the XML file.

What truly impressed me was its iterative process. Agentforce ran test cases, identified issues, and then suggested and implemented resolutions. It was like having a developer by my side constantly testing, refining, and writing code. After several rounds of this guided iteration, I had a complete solution, a modified Flow with the new Apex Classes (including a test class) that Agentforce had written and refined. The best part, I was able to push all of it, the updated Flow and the three Apex Classes, directly from VS Code into the org with one command. This project, which would have previously required a developer and significant back-and-forth, was completed by me, an Admin, thanks to Agentforce.

### Limitations and Considerations (Because It’s Not Perfect)

While Agentforce Vibes is incredibly powerful, we must approach it with a clear understanding of its role:

1. We still need understanding: a foundational understanding of what the code should do and how Salesforce works. When prompting, I always made sure Agentforce understood the scope of its actions, explicitly telling it not to touch other code and asking it to explain its proposed changes before execution.
2. It’s a tool, not a replacement for us: Agentforce is a sophisticated tool, not a substitute for our human intelligence. We must check its work. I treat its output as a highly intelligent first draft.
3. Code review is essential: Before deploying any AI-generated code to a production environment, a thorough code review by a developer colleague is essential. This ensures quality, adherence to best practices, and prevents any issues in production.

### Our Admin Work with Agentforce: The Future is Admin+

AI isn’t just transforming our Admin role; it’s elevating it. The Admin+ model, where we are Admins augmented by AI tools like Agentforce Vibes, is rapidly becoming the standard. Admins who embrace this technology aren’t just keeping up; we’re becoming more versatile, capable, and ultimately, more valuable members of our teams. We can contribute to projects that once required specialized developer resources, accelerating innovation and efficiency within our organizations.

### Conclusion

Agentforce Vibes significantly expands our capabilities as Admins, allowing us to undertake more complex and higher-skilled work than ever before. With this Salesforce native agent, we can confidently write, interpret, and debug Apex code, bridging that gap between declarative and programmatic solutions.

The narrative isn’t about AI replacing us Admins, it’s about AI empowering us, making us more valuable, more capable, and more integral to the success of our Salesforce implementations. The future for Salesforce Admins is bright, and I truly believe it’s powered by intelligent tools like Agentforce Vibes.


Agentforce Vibes is changing what it means to be a Salesforce Admin. For years, our work revolved around no-code and low-code tools—but now AI is opening the do

agentforce-vibes-can-we-really-use-ai-to-write-code-as-salesforce-admins

Agentforce Vibes: Can We Really Use AI to Write Code as Salesforce Admins?

AI is inescapable in 2026. Everyone is betting on it, and Salesforce certainly is with [Agentforce](https://www.salesforce.com/agentforce/). Salesforce first launched Agentforce at [Dreamforce](https://www.salesforce.com/dreamforce/) 24 to much fanfare, with integrations in Salesforce, Slack, and Mulesoft products. One of the big promises was adding autonomous AI to the platform, allowing it to work without input from staff, serving customers.

Since then, they have introduced 2.0, and at Dreamforce 25, Salesforce announced their newest version of Agentforce, with even more intelligence, better guardrails, and a [new voice mode](https://www.salesforce.com/agentforce/what-is-new/). In this article, we’ll take a look at how Agentforce has changed in the short time that it has been alive, and how realistic the promises Salesforce makes actually are.

## From Einstein to Agentforce: A Quick History

Agentforce isn’t Salesforce’s first venture into artificial intelligence. Salesforce’s first foray into AI was with the launch of Einstein in 2016. In its early days, Einstein offered limited but powerful capabilities, including predictive analytics, Lead Scoring, Opportunity Insights, and Forecasting.

By 2017, Salesforce expanded its AI offerings with Einstein Bots for Service Cloud, enabling automated responses to basic inquiries, intelligent escalation of complex cases, and record updates. The following year, 2018, saw the introduction of Einstein Prediction Builder, empowering administrators to create predictive models without coding, allowing them to forecast churn, identify upsell opportunities, and predict payment delays.

Salesforce leveled up its AI integration in 2023 with the introduction of Einstein GPT. This innovation combined the power of public LLMs with proprietary Salesforce CRM data, enabling users to generate custom and personalized emails, call summaries, marketing copy, and knowledge articles. This shift from purely predictive to generative AI was a significant game-changer.

However, generative AI’s true potential is unlocked through interaction, which is where Einstein Copilot, released in 2024, came into play. Copilot serves as the conversational layer, allowing users to interact with the AI, such as asking it to display opportunities based on specific criteria or generate comprehensive reports. Essentially, Einstein GPT acts as the brain, while Copilot provides the voice and hands, making it accessible and actionable.

Now, Agentforce is core to the Salesforce strategy. It is a key part of their messaging, even renaming the platform to Agentforce 360 and embedding Agentforce in everything from developer tools to user-specific solutions, as well as letting people build their own agents. If you’re confused by what Agentforce “is,” you are not alone, since the name is now on many of the Salesforce products. When determining your agent strategy, make sure you’re clear on your business goals, *then* choose the right set of tools.

### **Salesforce AI Evolution at a Glance**

| Layer | Einstein GPT | Einstein Copilot | Agentforce |
| --- | --- | --- | --- |
| Role | Generative AI | Conversation Assistant | Autonomous Agent Platform. |
| Launch | 2023 | 2024 | 2024 (1.0) |
| Focus | Produces content, predictions, insights | Natural-langauge, Q&amp;A, guided actions | Executes workflows, resolves cases, orchestrates across systems |
| User Interaction | Mostly behind the scenes | Direct, user-facing | Often background, monitored via Command Center |
| Strength | The brain | The interface | The operators |
| Example | Drafts a personalized sales email | Allows a rep to ask “draft a thank-you email” | Agent sends the email, updates the record, schedules a follow-up |

Yet, Salesforce wasn’t content with the user still being actively involved in every step. The vision was to go further: to have AI act autonomously, within defined guardrails, to execute insights automatically.

## **Inside the Agent Builder: How Agentforce Actually Works**

This ambition is precisely where Agentforce steps in, leveraging all the data within an organization to power its actions. So how do these agents actually work? The process begins in the Agentforce Builder, where the first step is to define the agent’s purpose. For instance, let’s consider creating a Sales agent to assist customers in purchasing generators at our hypothetical solar company. To get started, we would build a new topic, which we’ll simply call “Sales.”

![](/uploads/wordpress/2026/02/image-8-1024x785.png)

The Agent Builder is a fully-featured environment, allowing for the setting of detailed descriptions and precise scoping of the agent’s operational boundaries. In the initial setup, simply providing “Sales” as the topic prompted Agentforce to automatically complete almost all of the information needed. We can then add specific training data, such as details about generators, to further enhance its understanding.

Every step of this process is customizable, enabling you to fine-tune the agent to perform only what you define and to adhere to specified parameters strictly. This level of granular control is a core strength of Agentforce, ensuring that the autonomous AI functions not as an opaque “black box,” but as a carefully calibrated extension of your business logic.

The ability to customize every detail, from topics to the data it is trained on, to the actions it is allowed to take, is what makes it so powerful. With this new “Sales” Agent we have created, it can interact with a customer, taking their shipping information, answering product-based questions, and processing payments all from a chat window. This enables sales teams to increase volume and allows the highly skilled salespeople to maximize their productivity and profits in ways never seen before.

Setting up an Agent seems incredibly straightforward and downright, almost foolproof, but how do they perform in the real world? Let’s go ahead and use Salesforce’s example to see if an agent provided with the proper information can book me a dinner for two.

![](blob:https://cloudanswers.com/1f032f83-70e2-462b-9efb-3548a58ab1dc)

This agent is going to help our user Jim Halpert book a farm-to-table cooking class.

![](/uploads/wordpress/2026/02/image-9-1024x767.png)

As you can see, the agent is following its instructions by asking for both the member Id and email before proceeding with the conversation. Once I provided that information and said yes, it booked the class for me. However, as you can see, it is not perfect; it failed to provide the cost information to me after explicitly asking for it. The booking does successfully show on the customer record.

On the customer record, we are able to specify the Experience Interests, which Agentforce can use to recommend related events that it thinks the user will like.

![](/uploads/wordpress/2026/02/image-7-1024x920.png)

Using this information, let’s try another test with the Agent, seeing if it can recommend an experience for the customer and book it, using the customer’s interests.

![](/uploads/wordpress/2026/02/image-10-1024x675.png)

As you can see here, the Agent knows that the customer enjoys Cultural Tours and Workshops, Dining Experiences, and Spa &amp; Wellness. Using this information, it was able to suggest a handful of experiences that the customer would be interested in. Once I selected the Spa Retreat, it booked it for me and added that booking to the contact record.

Agentforce Agents are capable, but not perfect; they still need a lot of guidelines and training. As you saw from the first example, it did not pull in the price data, even though I had asked it to.

The reason it failed was that in the instructions, I did not tell it that it had to share pricing information if asked. It’s a great start; ideally, it would be smart enough to do that on its own, but situations like that are things to consider when building out and deploying your own autonomous agents. Another thing to consider is the cost of the agents.

Agentforce’s advertised cost is $125-$550 per user per month, on top of the Data Cloud and implementation costs, which puts you at 50k on the low end for existing users, and well into the 6 figures for new users.

Compare that to the possible cost of implementing your own solution, using APIs and prompt engineering to send data back and forth to an external LLM, and it makes you stop to think about whether the value is really there yet, and if it aligns with the cost.

## **The Verdict: Powerful, Promising, and Still Maturing**

Agentforce represents a genuine leap forward in what’s possible with autonomous AI in the enterprise. The Agent Builder is intuitive, the customization options are extensive, and when properly configured, agents can deliver impressive results.

But “properly configured” is doing a lot of heavy lifting in that sentence. These agents don’t think for themselves—they execute what you explicitly define. Miss an instruction, and you’ll have gaps in functionality and a poor user experience.

The pricing model has improved with Flex Credits, but costs can still climb quickly for enterprise deployments. Make sure you model realistic usage scenarios before committing.

Is Agentforce worth the hype? For the right use cases, absolutely. For others, it might still be an expensive work in progress. The key is understanding exactly what you need and testing thoroughly before going live.

Whether you’re evaluating Agentforce, optimizing your existing Salesforce environment, or need help building out automation,[CloudAnswers can help](/contact). We’ve been a Salesforce partner since 2008, helping clients get more out of their platform through[consulting, apps, and custom development](/products).


Salesforce’s vision for autonomous AI has been years in the making, but Agentforce 360 represents its boldest step yet. From defining agent behavior in the Buil

agentforce-worth-all-of-the-fanfare

Agentforce: Worth all of the Fanfare?

for-administrators

For Administrators


# Lazy Admin: A Great App for Quickly Querying Data

In our first app review, I’ve chosen an app that is simple, effective,
and works well: Lazy Admin by HTTPeak Pty Ltd. This application uses AI
and natural language processing to run SOQL queries and generate reports
and dashboards. You can ask it to query any data set in your Org,
including standard and custom objects.

## Querying Data

Let's walk through an example. I’ll have the AI query all accounts and
related contacts, and let's see how it does.

<img src="/uploads/lazy-admin-review/image2.png" />

I created a new search for all accounts and related contacts, sorting by
account number. It retrieved all records in the org, just as I asked,
with the related contact. The contact level in the report was included
as well.

## Exporting Data

Let's export this data via email so we can see it in a spreadsheet
editor. The email screen allows you to customize a message, select a
user to send it to and add external emails to CC so that you can send
the data out of the Org right from Salesforce.

Here’s what the email looks like with the custom message included.
<img src="/uploads/lazy-admin-review/image1.png" />

Here’s all the data dumped out, and it did exactly what it said it would
do. It got all Accounts, with related contacts and included the
relationship level. These were just the default fields. You can
add/remove fields either from the GUI or in the query editor, but for a
quick text prompt, it did a good job.

<img src="/uploads/lazy-admin-review/image3.png"/>

## Other Thoughts

There are some drawbacks to the application; you have the search history
and can favorite queries, but at the end of the day, they aren’t
Salesforce reports and can’t be saved and shared within Salesforce. I
had some security concerns, I reached out to the developer for comment
and this is what they had to say:

“We do not send the data out of the Salesforce org.However, we do send the metadata (schema) along with the search phrase, for better context. In addition to that, we have numerous orchestrations in place which are passed as system instructions to the LLM behind the scenes. This helps the LLM understand the appropriate context and instructions required to generate the SOQL.” - HTTPeak

Whether or not this is secure enough for you will depend on whether you
have any proprietary metadata. HTTPeak has passed Salesforce’s security
review and no record data leaves the org.

## Setup

The setup was rather easy. I went to their website and followed the
installation guide. The only limitation is it only works for lighting
apps, so if you don’t pay attention to that fact in the setup you may
run into an issue.

## Target Users

The users who would benefit from this are your lazy admins, who don’t
feel like writing SOQL queries, and any end user, say a sales manager
who already has access to all data, who just wants a quick answer on the
data without having to build out a report.

## Alternatives

Running a SOQL query from Salesforce Inspector, the developer console,
or for something permanent just a standard Salesforce report.

## Errors

I asked for all Opportunities related to Universities and Hotels, and
the AI was expecting to find record types. Instead, what I was looking
for were Opportunities with those words in the name. Understanding that
and rewording it got it to work, but things like that should be kept in
mind when interacting with the AI.

## Tips

Be sure to word your prompt as if you are talking to a computer. You
still have to keep that in mind. The LLM is good but imperfect, so
understand that it may make an error, and you might have to reword it or
edit the query it returns.

## Pricing

This app is a freemium app with three pricing tiers: free, $3,048 for
the starter tier, which removes the monthly limit and allows 12k
searches annually, and the Plus tier, at $6120 annually, allowing 60k
yearly searches and creation of dashboards. For free, you can run 10
unique queries a month, with no limits on how often you can send out the
CSVs via email or download them directly from those 10 queries.

## Conclusion

Overall, I’d say Lazy Admin is a solid application if you install it and
keep the users to just Admins/trusted persons at the free tier. In my
experience consulting, it’s nice to run a quick SOQL query without
having to type it out. It saves you from tabbing into ChatGPT, and it’s
a friendlier UI for reviewing data than other options like Salesforce
Inspector or the Developer Console. It’s a Lightning App, and it feels
and behaves like it. There is a fluidity to it, which is nice and
enjoyable to use. I like it more than Salesforce Inspector because it
generates my queries, and I can manually edit them if necessary.
However, I can count on one hand the number of times I run one-off
reports a month. Most reports are reused, so it’s more helpful to save
and share within Salesforce and sometimes email them out on a consistent
basis. I don’t see this replacing reporting, but for quick queries, this
is a great tool.

## Our Rating

**Functionality ⭐️⭐️⭐️⭐️⭐️**

**Innovation    ⭐️⭐️⭐️⭐️**

**UI and UX     ⭐️⭐️⭐️⭐️⭐️**

**Support and Documentation ⭐️⭐️⭐️**

**Security ⭐️⭐️⭐️⭐️**

**Value for Money ⭐️⭐️⭐️**

**Coolness ⭐️⭐️⭐️⭐️**

### **Overall** **⭐️⭐️⭐️⭐️**

[<u>View on the
AppExchange</u>](https://appexchange.salesforce.com/appxListingDetail?listingId=96ad3719-43d3-4ccd-a9ae-62cc933661ce&tab=r&utm_source=cloudanswers_website&utm_medium=blog&utm_campaign=lazy_admin_blog_post)


A Salesforce App review about an AI reporting application for the Salesforce Platform that runs your reports for you.

lazy-admin-review

AppExchange Review: Lazy Admin

Salesforce apps, powerful components, development and consulting. Our experienced team helps you to create and modify workflow processes in salesforce.

Ai

Agentforce Vibes: Can We Really Use AI to Write Code as Salesforce Admins?

Agentforce: Worth all of the Fanfare?

AppExchange Review: Lazy Admin

Partners

Company

Get in touch