You Were Hired for a Different Job Than the One You're About to Have
Your platform evolved. Your team didn't.
The person running your CRM is probably very good at their job. That’s not the problem. The problem is that the job is becoming something else — and the skills that made them excellent at the old version don’t automatically transfer to the new one.
They were hired to manage email flows and segmentation. A/B test subject lines. Build a welcome series. Suppress unengaged contacts before a big send.
I’d wager that within 18 months, that same platform will be the layer where your brand’s AI agent and your customer’s AI agent negotiate recommendations, manage cart logic, and close transactions without a human touching the keyboard on either side.
Those are not the same job.
This isn’t a threat piece. It’s an opportunity map. The gap is real. It’s large. And it’s closing faster than anyone is preparing for. The marketer who closes it first isn’t in trouble — they’re in an extremely good position.
What Your CRM Is Actually Becoming
Your customer doesn’t want to talk to you anymore — not to your chatbot, not to your support queue, not to your drip campaign. They want to send their agent to deal with yours. I wrote about that shift in January. This piece is about what it means for the infrastructure sitting at the center of your marketing operation.
We’ve been calling it a CRM — customer relationship management — for so long that the name has become a liability. It implies a tool for managing relationships between humans. A place to store contacts. A mechanism for sending scheduled messages to segmented lists.
That’s what it was. It is not what it’s becoming.
The shift underway is from CRM-as-messaging-tool to CRM-as-SSOT: Single Source of Truth. Not a platform that humans query through a dashboard to build campaigns, but a structured data layer that AI agents query in real time to negotiate, transact, and personalize at machine speed. IDC projects that by 2026, nearly half of new CRM investment will go to data architecture and AI infrastructure rather than additional licenses or modules. The money is following the architecture.
The primary user of your CRM used to be a marketing specialist scheduling sends. It is becoming an autonomous AI agent pulling live signals to make decisions that used to take a human three clicks and a gut check.
Relational databases optimized for dashboard readability are giving way to structured endpoints designed for machine access. Scheduled blasts are giving way to real-time exchanges between brand-side and consumer-side agents — adjusting offers, handling objections, closing transactions in the time it used to take your automation to fire a trigger email. And the success metric that matters is no longer open rate or click rate.
The metric you need to be measuring is inference advantage — how much better your agent performs because it has richer, more current, more accurate data than the agent on the other side. That’s not a metric most lifecycle marketers have ever been asked to think about, let alone optimize.
Two platforms are worth naming because the specifics matter.
Klaviyo moved first and moved fast. In August 2025, they launched an MCP server — a standardization layer that allows third-party AI agents to connect directly to your Klaviyo instance. Claude, ChatGPT, whatever agent stack you’re running — they can now query your performance metrics, segments, and campaign history using natural language. No custom integration. No API connector built by your dev team. One standardized interface. They also shipped a Customer Agent inside K:Service — a brand-side AI that handles customer interactions across chat, SMS, and email, with full access to your first-party data and personalization logic.
Salesforce is building toward the same model with Agentforce, though they’ve had a rougher ride. In September 2025, Noma Security disclosed ForcedLeak (CVSS 9.4) — a prompt injection vulnerability that let attackers embed malicious instructions in Web-to-Lead forms, causing AI agents to exfiltrate sensitive CRM data. The security architecture for agentic interaction is still being built in real time.
The underlying protocol structure is what changes the scale equation. Before standards like MCP existed, every agent needed a custom connection to every data source — the classic M×N integration problem. MCP reduces that to M+N: implement once on each side, be accessible to every major agent ecosystem. Layer in the Agentic Commerce Protocol (ACP, maintained by OpenAI and Stripe) and Google’s AP2 for payment authorization, and you’re looking at a fundamentally different architecture for how brand data moves through the world.
The Skills Gap
Here’s what I keep thinking about.
The lifecycle marketer who has spent five years getting very good at segmentation strategy, flow logic, subject line testing, send-time optimization, and suppression management has built a genuinely valuable skill set. None of that was wasted. But that skill set was built for a platform that delivered messages to human inboxes and measured whether humans opened them.
The skill set that the same platform now requires looks different:
Protocol negotiation* Understanding how MCP endpoints are structured, what agents can and can’t access, and why that matters for what your brand can offer at the moment of intent.
Agent authentication* understanding how your brand-side agent proves it has authorization to act, and what happens when it doesn’t.
Inference optimization* structuring your first-party data so your agent has a material advantage in a negotiation with a consumer’s agent that has its own data signals.
Data governance as a real-time problem* not a quarterly hygiene exercise.
These are not impossible skills to develop. (They’re not even that exotic — a marketer with strong technical instincts and intellectual curiosity can get there.) But they are not what your current team was hired for, and they are not what most lifecycle marketers are building toward right now.
The organizations that close this gap first won’t be the ones that replace their marketing teams with engineers. They’ll be the ones that identify which marketers have the instinct for this kind of technical fluency, give them room to develop it, and build roles that didn’t exist two years ago.
The Security Layer Nobody Has Staffed For
The ForcedLeak vulnerability exposed something the whole industry will have to reckon with: agent-to-agent communication creates attack surfaces that traditional marketing security models weren’t built to handle. When your brand’s agent is processing data through a shared protocol layer, the question of what gets exposed to whom is not a simple one.
This used to be a security team problem. It is becoming a lifecycle marketing problem — because the people who know your customer data most intimately, how it’s structured, what’s used for personalization, what’s sensitive — sit on the marketing side. And nobody has asked them yet which of that data should be queryable by an external agent. (I’m going deeper on this in the next piece. For now: if nobody on the marketing side has been in a room with security to discuss protocol-level data exposure, that conversation is very long overdue.)
The Career Opportunity
Gaps are where careers are made. Not gradually, through steady advancement — but in the moment when everyone else is still reading the old map.
The shift from CRM-as-email-tool to CRM-as-agentic-SSOT is creating that gap right now. The platforms are ahead of the practitioners. The infrastructure is moving faster than the job descriptions. The roles that will exist in 2028 — growth engineer, agent orchestration strategist, agentic data architect, lifecycle protocol manager, whatever we end up calling them — are not common yet. But the people who will fill them are already working somewhere, probably debugging why a segment isn’t behaving the way it should.
The marketers who get there first won’t be the ones who waited for a job description to tell them what to learn. They’ll be the ones who looked at what their platform was becoming and started asking: what would I need to understand to be useful in that world?
Protocol architecture. Agent authentication. First-party data governance. Inference advantage. These aren’t developer skills — they’re strategic skills, and they sit exactly at the intersection of marketing judgment and technical fluency that good lifecycle marketers already operate in.
The job you were hired for may not exist in its current form for much longer. The platform does. The question is whether you’re building toward what it’s becoming.
*Up next in the Agentic Shift series: The transaction — what commerce looks like when neither party is human, and why the marketer’s role shifts from campaign creator to agent trainer.
