AI in Action is brought to you by Fast Slow Motion. Our team helps growing businesses put AI to work with practical, scalable solutions. To learn more about how we can help you implement AI in your business, visit fastslowmotion.com/AI.

Welcome back to another episode of AI in Action. I’m Eric Housh with Fast Slow Motion director of AI Zack Terry, Salesforce Winter 26. Just landed and it’s loaded with AI. In this episode, we’re going to boil it down for you to find out what you can use now and what’s worth planning for next quarter. That’s right. Yeah, we really are seeing so much AI, which of course we expect with any of these Salesforce releases and anything that they’re putting out, AI is kind of central to everything.

And so I think the the key here is they are pushing AI into more of the places that your teams already work. So a lot of focus on sales, marketing, getting clear insights, allowing admins to get AI inside of Flow, and leaders to get easier pilots with Agentforce. So definitely a lot to talk about. And we’re going to get to that in just a few minutes.

But like we always do, Zack, tell us what’s going on. The biggest developments in the AI landscape over the last week. It was a big week, a lot of announcements. So I’m going to blitz through these. And if you’re interested in learning more, of course, go and check out some of the articles that refer to these. But first up we have open AI.

We had a really, really big week with open AI. A lot of new announcements and releases coming from the AI team over there. So first of all, we have Codex moving from a research preview to general availability. They’re bringing a full SDK. They’re bringing Slack integration admin controls, all powered by GPT five Codex. There’s a new Apps SDK that allows developers to create apps directly on ChatGPT.

So really turning ChatGPT into a platform and this is competing with a lot of startups and a lot of companies that are building apps that integrate with ChatGPT. Well, now you can build them directly on the platform and access them from ChatGPT, which is pretty exciting. The demo that they showed was very cool. They’re adding something called AgentKit, which adds infrastructure and evaluation tooling so that organizations can build and deploy their own AI agents.

So together, all of these launches really move OpenAI from a model provider to a full on agent platform. So a lot of capabilities that they’re bringing that I think a lot of other competitors or other businesses have been building around, and now we’re essentially able to build that directly on top of OpenAI. So not only do we have the ability to build apps, but AgentKit is essentially a graphical interface where you can build workFlows and create these sort of different agent nodes.

So very interesting. I’m excited to see what people build with this. One thing to keep in mind if you are building with this is Everything Uses tokens. So even if you’re going to be working in that workFlow, you could be, you know, building nodes that don’t necessarily rely on an LLM for generating some kind of output, but everything’s running through that engine.

And so the entire process is going to consume tokens. So just something to keep in mind where it may be more expensive to develop on open AI. All right. So second, we have Google launching Gemini Enterprise for AI agents at work. This is a unified platform for conversational access to company data that comes with some pre-built and custom agents.

It’s being priced at about $30 per user per month for enterprises. But if you’re a small business, you can get that brought down to $21, which is nice to see a little bit of a discount there. And ultimately, this is sort of competing head to head with copilot on 365 and ChatGPT enterprise with their enterprise plan. And so it’s really focusing on workFlow automation and hopefully bringing some clarity on that return on investment.

If you’re investing in these AI tools. Again with open AI, for our third item here, they have tapped AMD for multi gigawatt AI chip supply. So they’ve signed a multi-year deal to provide up to six gigawatts of AMD instinct GPUs, with the first gigawatt set to deploy in 2026. And it includes warrants for up to 160 million and shares, which is about a 10% stake in the company.

So this is really yeah, it’s it’s wild. More investment coming in and really diversifying OpenAI’s compute supply beyond Nvidia and strengthening those hardware economies for, you know, more large scale AI workloads. All right. So two more to hit here Eric. Two more items. One yeah that’s right. So many things all at once. So very interesting here. Deloitte has rolled out Claude or is planning to roll out Claude across 470,000 employees, which is one of the largest enterprise AI deployments we have seen so far.

So that’s a massive workforce being able to use this tool across the enterprise and as a part of this, they are creating a cloud center of excellence and certifying 15,000 professionals as internal AI specialists. So this is really some mainstream adoption of AI assistants inside a professional services firm, and not just a professional services firm, one of the largest in the entire world.

It’s Deloitte. Exactly. So seeing this sort of adoption at such a large firm is pretty, pretty interesting. I wonder if we’ll see other large firms follow suit and and create similar processes. And I like this idea of a center of excellence, you know, maintaining how we’re using that tool across the organization, putting some checks and balances in there.

I think it’s overall, it’s a good idea. All right. And finally, we have the European Union launching a €1 billion strategy called apply AI. So this is public sector funding to drive AI deployment across health care, energy, automotive, manufacturing and defense spaces. It’s really kind of similar to what the United States is doing with the AI in Action plan or the AI action plan, not the mix.

Our our podcast name here with the infrastructure bill. Yeah, they totally stolen it. Exactly, exactly. Yeah. We were we were the first. We were definitely the most original. Right. But, ultimately it’s very similar to aiming to boost European sovereignty and close that adoption gap between US and China and sort of all the infrastructure investment that’s going into AI there.

So, it should open up some new procurement pathways and incentives for enterprise AI rollouts across the EU. So I think it’s interesting. It’s really just a continuation of what we’re seeing where these, very large scale governments and entire nation states are putting some strategies around AI. And I think it’s really sort of leaning credence to this idea of this AI race.

It’s sort of like the race to the moon. It’s it’s we’re all trying to develop these technologies as fast as we can so that we have an advantage. And so I think this is just another step where we have the European Union putting a lot of money towards making sure AI gets rolled out effectively. It feels a little bit like an arms race.

Absolutely. But thanks for running through all those headlines. With us, as always, if you want to dig deeper on that, we’ll have links to those articles in the show notes. You can, get into, but let’s shift gears. Let’s talk about Salesforce’s Winter 26 release. As of the date we’re recording this, it is fully out there, and it is loaded with AI.

Let’s start with Agentforce, what’s real and what’s worth trying. So two big sort of initial Agentforce wins that we’re going to see with this release. One is related to the SDR agent. So if you have tested the SDR agent, maybe you’ve rolled one out yourself across your business. Maybe you’ve run into a conversation where you needed to have more than one, but you couldn’t do that because you could only use a single SDR agent.

So basically you define a purpose and it goes out and it executes on that purpose. But if you needed to use the same functionality for something a little bit different, well, you, you essentially you either couldn’t or you had to repurpose that existing agent. Well, now Salesforce is bringing the ability to have up to five different SDR agents in a single org, which is great.

So you could essentially spin this up for different ICPs or different sales motions. You can use different prompts, different rules. You could A/B tests different SDR agents to see you know which one is performing better. You get higher daily email limits sort of across the board. So I think they’re bringing that up to 10,000 emails per day now, which is quite a big job, especially if you were on Gmail, because that was formerly limited to 1800 emails a day.

So you’re getting, you know, more than five times the number of daily email sends. If that was what you were using. So I think this this really just gives us the ability to experiment more. It allows us to really track a whole lot more meetings that are being booked by these SDR agents across the board. So this is a real problem that I think a lot of people felt with the SDR agent, and to see it addressed so quickly, I think is really fantastic.

And ultimately I think it’ll lead to higher adoption of this tool. Yeah. So then the other thing is, is a lot of selling aides using AI are being brought onto the platform. So last year if you watched or attended Dreamforce, you probably saw this account summary tool where you were able to essentially just look at an account record or look at an account plan and get a bunch of generated insights about that account.

And that was the last we saw that. Well, now it’s finally coming into production. So this is going to be generally available. You’ll be able to add the summary to accounts to account plans. It’ll be able to go out and perform research on your behalf. You can compile public information. And public signals, find competitive notes, essentially just helping sellers prep even faster.

Or for account managers, people that are working on client retention allows you to maybe spot some signals, get ahead of some things so that you can have those conversations with your accounts. So I think that’s a really great opportunity. And then we’re also going to see more summary capabilities across the sales organization. So summaries for accounts contacts leads opportunities a lot of the stuff we’ve seen on the service side of the house, we’re now seeing being brought over to the sales side as well.

And we can share a bit about marketing. Of course on the marketing side, Agentforce now able to pull campaign insights so you can ask the agent for performance, get a clear readout and then inform sales kind of what the next step is really closing that loop. Absolutely. Yeah. Really great to be able to just ask an agent, hey, like what’s going on with this campaign or how are we doing?

Where are we looking at from the budget and the ROI? So it’s really I don’t think it’s anything crazy. We’re just tying in the capability for the agent to go out and retrieve those insights from the marketing side of the house. But if you’re already working inside of Agentforce and you’re using that day to day, it’s just a great addition to be able to layer in that marketing information.

Let’s talk about Flow Flow, obviously the automation backbone of Salesforce. What has changed in this release for admins who want AI without having to rebuild everything? So what I will say is, as with any Salesforce release Flow, at least for the past 4 or 5 years, has been a huge focus. So what we’ll talk about today is just a very, very small amount of the overall updates that have been made.

So definitely check out our podcast episode on the release if you want to learn more, go more in depth, check out the release notes or any of the information that we’re linking in the show notes here. But for the AI side of things, we are getting the ability to use AI inside of a Flow decision element. So instead of building a multi branch sort of logic tree within a Flow.

So let’s let’s say for some reason you had ten different branches of logic where you got a decision and it’s making a decision on which path you should follow. And you’ve got ten different variations. Well, instead of having to go in and actually programmatically design what each and every one of those variations are, and defining the variables and what we’re looking for, we can now do that in plain language and just describe that to the AI, and then it will create those decision paths for us.

So it’s able to go and essentially just make that whole process a little bit easier. Now of course with AI, you always got to make sure that you are validating and auditing and making sure that it’s it’s creating the logic that you expect. But to be able to use AI to create some kind of decisioning inside of a logical process is really useful beyond just the shortcut, because it could also be something that’s maybe a little bit ambiguous.

So maybe you’re getting some free text as an input and you’re not exactly sure what you need to be looking for. You can use AI to kind of make that judgment and ultimately try to push you down the right path. So of course, the caveat always keep a human in the loop. Always make sure you can audit and check what it’s doing.

But I really do like the ability to bring that natural language directly inside of Flow. Yeah, seems super powerful for sure. Yes. Agreed. And then one thing to keep in mind you do have to have the correct entitlements. And so you do have to have Agentforce in your org. If you’re not an Agentforce or you won’t be able to use the AI in decision elements.

So do keep that in mind. If it’s something you’re interested in and you don’t have Agentforce, make sure to to either you know, get on foundations, get that free skew to test it out. But if you want to take advantage of this, you will need to have that in your work. And then we’ve got a couple of just kind of quick hits for for some other Flow updates.

Create records are now returning fields immediately. So that means you don’t have to have a separate get element to go and retrieve the information that you just created, which is a nice quality of life update. There’s the ability to compare versions across Flows now, so you can see differences and track changes. There’s an improved debugging window. There’s centralized test execution that’s combining Flow and apex into a single window where you can run all of your tests, which is really nice.

And then we’re getting more Flow analytics and a brand new time-based automations view, which I think is funny because for a very long time it’s been called time-based workFlow, which if you know what’s been happening in Salesforce for the past several years, workFlow rules, process builder, those are essentially dead in the water now. Right? And so they’re really making some nomenclature updates to kind of get the word workFlow out of the system.

Salesforce changing the names of things. That doesn’t sound familiar. Never heard of such a thing. So bottom line if you’re an admin grab one of those messy Flows. Refactor it with that AI decision element, measure handle time. Measure error rate. Keep yourself in the loop. Make sure it’s behaving as expected and I think you really feel some impact there.

Let’s talk about conversation intelligence and Data Cloud call intelligence and Salesforce seems to just be continuously improving. What’s new with this release? The big update here is you can now plug this information into Data Cloud. And that sounds kind of small, but considering that if you want to be able to leverage semantic search and search through those conversations, that’s essentially that’s powered by Data Cloud, because it has that vector database where we can take that information and we can create the embeddings and create the search index.

And so that wasn’t there before. But now that we have the ability to just plug that directly into Data Cloud, that means that we can get summaries much quicker. We can mine that information and search through that information much easier because we have the ability to do those semantic queries. So ultimately it can lead to better coaching, mining more information, finding objectives, finding next steps, basically being able to see patterns in those conversations without having to listen through every single call.

But the main update there is the Data Cloud empowerment, being able to leverage semantic search within those conversations. Let’s talk for a little bit about knowledge. What is in this release that’s really going to make life easier for those support leaders? For anyone that’s using knowledge in Salesforce, I think this will be interesting because there’s a new feature called generative Q and A that is specifically for knowledge articles, and what it can do is scan the content of your articles and then it can propose question and answer blocks.

So let’s say you have a really long form piece of content. And you wanted to create something to make it easier to skim through. Maybe just highlight the key questions that people might ask with some quick answers to those questions. Generative AI can now basically scan through the document and try to create that for you, so it can create the questions and answers that it thinks people will most likely ask when they read the article, which is nice, because this is something that I think a lot of people do anyway.

It makes it a little bit easier to understand the articles, but what can happen now is when you’re generating those questions and answers, it can rank common questions, and it leads to the ability to author these articles faster, create sort of tighter documentation around it. Basically just a easy way to prove some AI ROI, because it should be reducing the overall time that you would spend writing those articles or having to create the summaries.

Now it is kind of a niche update. You know, if this is not something that you would need, of course, I don’t know that you would get a lot of benefit from it. But if you use a lot of knowledge articles and you’re always looking for ways to make them easier to read, easier to understand, this is a great way to do that.

And then also the side effect here is if you’re using your knowledge articles with generative AI, let’s say to get customer answers from the service side of the house. Or maybe it’s sales questions, whatever that could be by having some Q&A, formatting and some of these articles, it may lead to better retrieval when you’re looking for those answers, because it’s going to essentially group that information into a way that’s going to be easier for the AI to understand and draw the connection between a question and ultimately the answer that it should retrieve.

So I think that’s a nice side effect that you might get by using something like this. Let’s talk on-platform Marketing Cloud. We got some upgrades there as well. Yes. And I think I would be doing everyone a disservice if I didn’t describe what we mean by on-platform Marketing Cloud. Yeah. So this this has been a little bit confusing because Salesforce has been not only creating their own marketing products, but they also continue to support the acquired marketing products that they have.

And so when I say acquired marketing products, I’m referring to Marketing Cloud, which is formerly called Exact Target, and Marketing Cloud Account Engagement, which is formerly called Pardot. So those are two separate databases that integrate with Salesforce. So those were acquired products that Salesforce is integrated into the platform. Well, now we also have products called Marketing Cloud Growth, Marketing Cloud Advanced, and Marketing Cloud Next.

And these are all of the marketing products that Salesforce has been building on-platform. So in other words, instead of acquiring a product and then bolting it on, this is being built from the ground up on the Salesforce platform. And so it’s been happening for a couple of years now. And what we’ve seen is features continue to be built and launched on top of this to ultimately bring it to parity with the other marketing tools.

So now we have a lot more power in these tools than we did when they were first announced. And I think ultimately, this is where the future of all of Salesforce marketing is going. This is where all the AI is being layered in. So if you wanted to use AI within your marketing stack, it’s going to be more feasible to do so if you’re using one of these on-platform tools, as opposed to using something like Marketing Cloud or Marketing Cloud account engagement.

So all that to say, just provide some clarity on what we are referring to with on-platform Marketing Cloud. Those would be growth advanced and next it’s the tools that are built on top of Salesforce. So all that said, Salesforce is continuing to consolidate AI features. So now you can have AI agents help you draft and revise content and not only help you do that, but it can also do it in your brand voice.

So this is something HubSpot’s had for a while where it can help you draft content and reference your brand voice and brand information. And now we’re seeing that capability come to the Salesforce platform as well. So that’s a great quality of life upgrade to be able to easily draft emails, create landing pages, create that content in your brand voice.

And we’re also going to have these reusable content blocks that we’ll be able to create once. And then reuse kind of throughout the system, different assets that we would have in our marketing tools. We have sandbox testing that has now come to these tools, which was previously lacking. So now we have the ability to test these updates inside of a sandbox.

So ultimately it’s continuing to just layer these AI features in where we think people are working and where we think that it can help. And hopefully bringing some tighter feedback loops between what we’re shipping and what we’re learning. Yeah, makes a lot of sense for sure. Let’s, speak to the geeks for a little bit. What’s in it for the developers with this release, I definitely had to get our director of development to walk me through what this meant.

Like a different language. But we’re going to get to. Yeah, I just wasn’t. I was like, I don’t know what this means. So, fortunately, we have a very smart director of development, and this feature is called Trusted Mode for Lightning Web Security. And this essentially opens the door for more third party web libraries inside of Lightning Web Components.

So if you are building lightning Web components, this should make it much easier to use third party libraries, which you know, apparently was a really big limitation around any sort of development that you might be doing with LWC. So another example would be if you had AI, and that depended upon a browser API and Locker may have blocked that.

Well, this this may not matter anymore, right? You may be able to actually access those libraries without having to, you know, create some workarounds or anything like that. So ultimately it’s speeding up delivery around those specific problems and reducing the amount of workarounds that you would have to create in order to make those things work, because it’s essentially bringing that trusted capability directly to where you would build so that all of that information super helpful, maybe a lot to consume for our audience here.

So let’s land the plane this week. Give our audience four quick actions they can do right now digging into these AI features in the Winter 26 release. Well, I’ll give you five because number one is just get Agentforce in some capacity so that you can start playing around with some of these tools. So do that through foundations.

If you don’t have it already or have a conversation with your AG, get that rolled out so that you can start playing with it. But specifically for things that you can start playing with. SDR agent, set up an SDR agent, set up multiple SDR agents. If you want to take 2 or 3 of your ISP’s and create SDR agents around them and measure how many net new meetings you’re getting, do some AB testing, see which ones are performing better.

Second thing you can do is go and add Einstein summaries to your sales objects. So accounts, contacts, leads, opportunities. Go play with the Account Research component. Add that to your account. See how that works. See what kind of questions that you might have and what’s able to go and answer. Really easy way to do that. Just drop it on the account layout, drop it on the account planning layout.

Make sure you sign the permissions, go in there and just kind of see how that works. Another thing you can do is take advantage of these AI decision element inside of Flow. So maybe go and take a Flow that you already have that uses decision logic and go ahead and refactor that and use AI instead and just kind of see how it works, right?

Ultimately, just experiment with it. I’m not saying go and activate it and launch that into production immediately, but go create a new version of it, test it out in a sandbox, just kind of see how it works, track any errors you run into, see if it’s actually doing what you want. Great way to just go and test how that works.

Then finally, if you have access to conversation insights, then definitely go and experiment with ingesting that information in the Data Cloud. Now obviously make sure you talk to your AE, understand the consumption model, make sure you’re only ingesting the information that that you know that you want to have. I believe it should be first party data since it’s coming from Salesforce.

And Salesforce made an update a few months back with Data Cloud. So any of that kind of first party data coming into your CRM is no longer being docked credit consumption. So I think that should fall underneath that. But definitely make sure to check with your AG if there’s any concerns there. But if you have it, then a great thing to do is to go ahead and ingest that conversation data into Data Cloud and then start testing that semantic search.

See what your agents are able to run through on each of those conversations, what insights you’re able to mine from those, maybe even hook that up to an employee agent that you might have running, and allow it access to those conversations so that you can get that information. So winter 26. It’s all about working smarter with AI, automating more of a Flow, moving faster across all the divisions sales, service, marketing, and so on.

It can be overwhelming. So if you’re listening to this, you need some help mapping these features to your processes. That’s what we’re here for. So head to fast slow motion.com/ai and get in touch. That’s right. And as a reminder just start small. Tie it to a metric and ship it. The value really shows up when AI is living inside the work that your team is already doing.

That great show. That’s it for this week’s episode of AI and actions. We’ll see you all next time.