Are your CRM workflows actually automated, or are they just scheduled? That’s the question worth sitting with, because the answer increasingly determines whether your go-to-market team is spending time on deals or on data entry.
AI agents for CRM automation are moving from experimental to operational for a growing number of sales and RevOps teams. They don’t just trigger pre-set actions based on field changes. They reason, route, and act across multiple connected applications in real time – pulling context from your CRM, your inbox, your Slack, and wherever else your customer data actually lives.
Two developments this week make that shift concrete. The Python package viaSocket hit version 0.1.2 on PyPI, expanding its library of connected apps to over 2,300 – including HubSpot, Slack, Google Sheets, and Notion – and making them available as callable tools for AI agent frameworks like LangChain, CrewAI, and LlamaIndex. Separately, Salesforce is reportedly acquiring a defense tech startup, extending a run of acquisitions that signals the company’s appetite for broadening what its platform can sense and act on. These aren’t unrelated stories. They’re two sides of the same shift in how CRM systems are being built and extended.
What AI Agents Actually Do Inside a CRM
The clearest way to understand an AI agent in a CRM context is to contrast it with a traditional automation rule. A rule says: if a deal moves to stage three, send a follow-up email. An agent says: review the deal history, check whether a follow-up has already been sent via a different channel, assess the contact’s recent activity, and decide whether an email, a Slack message to the account owner, or nothing is the right next action.
That’s a meaningful difference. Traditional automation optimizes repetitive tasks. Agents handle judgment calls that used to require a human to manually check three or four different tools.
For RevOps teams, the practical implication is that your sales pipeline stops being a passive record of what happened and starts becoming something closer to an active participant in what happens next. Fields get updated because an agent noticed a change in email sentiment. Deals get flagged because the agent cross-referenced the close date against the contact’s calendar availability. That kind of contextual awareness has been technically possible for years – what’s changed is how easy it’s becoming to wire it up.
Why viaSocket’s App Library Matters for GTM Teams
Most CRM teams aren’t building their own AI agents from scratch. They’re using frameworks – and those frameworks need access to the apps their users are already connected to. That’s exactly what viaSocket is designed to solve.
The library’s 2,300+ app integrations become available as tools that an agent built in LangChain or CrewAI can invoke directly on behalf of a specific user. It’s not a generic connector. Each user’s authenticated connections – their own HubSpot instance, their own Google Sheets, their own Notion workspace – are exposed as agent-callable functions. That means an AI agent can act with the permissions of the user, within the context of their actual data, without requiring a custom integration for every application.
For GTM teams, this matters for a few specific reasons:
- Sales reps work across five or more tools in a single deal cycle. An agent that can read and write across all of them reduces the friction of keeping data consistent.
- RevOps teams building internal automations no longer need to maintain bespoke API wrappers for every tool in the stack.
- Teams using HubSpot can give an agent the ability to create contacts, log activities, and update deal stages without building a full HubSpot integration from scratch.
The fancy-hubspot package, also updated on PyPI this week, takes a similar low-dependency approach – offering plain HTTP access to HubSpot operations without wrapping everything in a vendor SDK. It’s a small signal of a broader trend: developers building CRM-adjacent tooling are choosing lightweight, composable approaches over heavy platform SDKs. That makes it easier to embed HubSpot operations inside an agent workflow without inheriting a lot of unnecessary abstraction.
Salesforce’s Acquisition Strategy and What It Suggests
The reported Salesforce acquisition of a defense tech startup is harder to read in isolation – the details are thin. But it fits a pattern worth paying attention to.
Salesforce has been acquiring steadily, and the targets haven’t always been obvious CRM adjacencies. Defense tech, in particular, involves capabilities around real-time data processing, secure communications infrastructure, and complex workflow orchestration under constrained conditions. Those capabilities translate. A system built to manage mission-critical decisions under uncertainty isn’t far removed from what an AI agent needs to do when it’s routing a high-value deal through a complicated approval chain.
What it suggests for the broader market is that competition in CRM isn’t just about features on a pricing page anymore. It’s about which platforms can support the kind of agentic, multi-system reasoning that enterprise sales cycles actually require. For teams evaluating their stack, it’s worth checking our CRM Tools Directory to see how the major platforms are evolving on this front.
How This Changes the Way You Should Think About Your ICP
Here’s a less obvious implication. When AI agents can act on behalf of your users across dozens of connected apps, the definition of what your product can do for a customer starts to change – and that changes how you think about your Ideal Customer Profile.
A team that has deeply integrated their CRM with ten other tools will get dramatically more value from an agent-powered workflow than a team still copy-pasting between spreadsheets. That integration depth becomes a qualifying signal. It’s not just about company size or industry anymore – it’s about how connected a prospect’s operational stack already is.
This has downstream effects on Customer Acquisition Cost too. Customers who can immediately plug AI agents into their existing connected apps don’t need the same onboarding investment and start getting value faster. That compresses the time-to-value curve in a way that matters for Customer Lifetime Value calculations – teams that get faster value tend to stick around longer and expand more predictably.
What Email Marketing Agencies Reveal About Where Execution Is Heading
It might seem odd to bring email marketing agencies into a conversation about AI agents. Bear with this for a moment.
The emergence of a recognized tier of specialist email marketing agencies – including Klaviyo specialists and CRM creative studios operating across markets like London – reflects something real about where execution complexity is concentrating. As CRM platforms get more powerful, the execution layer gets more demanding. Segmentation logic gets more sophisticated, personalization goes deeper, and campaign sequencing becomes more intricate.
AI agents are going to accelerate that dynamic. They’ll handle more of the operational legwork – syncing data, triggering sends, logging responses – and push the remaining human work further up the value chain toward strategy, creative judgment, and audience understanding. The agencies thriving in 2027 won’t be the ones who mastered the mechanics of a platform. They’ll be the ones who mastered how to direct agents effectively.
For in-house teams, the same logic applies. The win rate advantage isn’t going to come from having agents. It’s going to come from knowing what to ask them to do and how to evaluate whether they’re doing it well.
Practical Steps for CRM and RevOps Teams Right Now
If you’re trying to figure out where to start with AI agents in your CRM workflow, the honest answer is: start narrower than you think you need to. Don’t begin with a fully autonomous pipeline manager. Begin with one agent that does one thing well.
- Identify your highest-frequency manual task. If your reps are spending time every day copying data between HubSpot and a spreadsheet, that’s your candidate. That’s a workflow an agent can own.
- Choose an agent framework that connects to your existing stack. LangChain, CrewAI, and LlamaIndex all have different strengths. The right one depends on how your team wants to build and maintain agents over time.
- Use lightweight integration libraries where possible. Tools like viaSocket and fancy-hubspot reduce the overhead of giving agents access to your CRM. You don’t need a full platform integration to get started.
- Define what “done well” looks like before you deploy. An agent that updates deal stages with 80% accuracy sounds useful until it corrupts your sales forecast. Set evaluation criteria first.
- Monitor with a RevOps lens. Treat the agent like a new hire on a 30-day review. Check its outputs against expected behavior weekly until you trust it on the tasks you’ve given it.
For broader context on how CRM tools are evolving to support this kind of automation, the CRM Guides section has practical walkthroughs worth reviewing before you commit to a build-vs-buy decision.
The Quiet Part Most Teams Are Missing
There’s a tendency to frame AI agents as a future capability – something to plan for next year. That framing is getting expensive. The teams building agent-powered workflows now are compressing their sales cycles and improving data quality in ways that compound over time. The gap between those teams and the ones still debating whether to start is widening faster than most people realize.
The release of viaSocket 0.1.2 won’t make headlines outside developer circles. But it represents exactly the kind of quiet infrastructure building that makes the next wave of CRM automation possible – 2,300+ apps, each of them now callable by an AI agent acting on behalf of your users, inside your workflows, against your actual data.
So back to the question from the start: are your CRM workflows actually automated, or are they just scheduled? If the honest answer is the latter, the tools to change that are already in the package registry. The remaining question is whether your team is going to use them before your competitors do. Stay current with what’s shipping by subscribing to the CRM Daily Newsletter – the signal-to-noise ratio on this topic is getting harder to manage alone.