The Signal
IBM reports that 2026 is the year of the "super agent" — AI systems that plan, call tools, and complete complex tasks across your browser, inbox, and business systems without you managing a dozen separate tools. One prediction making the rounds: AI agents will replace 50% of the SaaS tools businesses use today. Whether that timeline is right or not, the direction is clear.
The Story: What AI agents actually are (without the hype)
Every week I hear from business owners confused by the term "AI agent." Fair enough. It gets used to describe everything from a fancy chatbot to a fully autonomous digital employee. Let me break it down into three categories that actually matter.
Chatbots are the simplest version. You type a question, it gives you an answer. Think of the help widget on a website that tries to resolve your issue before connecting you to a real person. Chatbots follow a script, but they do not take actions on your behalf. They talk. That is it.
Copilots sit next to you while you work. They suggest email replies, summarise documents, or draft a first version of a report. The key distinction: a copilot does not act alone. It makes suggestions. You decide, you click, you approve. You are still in the driver's seat.
Agents are different. An agent plans a sequence of steps, accesses your data, uses tools, and takes action — often across multiple systems — with minimal intervention. Instead of suggesting a reply to an email, an agent reads the email, checks your CRM for the customer's history, drafts a personalised response, and schedules a follow-up. It does not just think. It does.
That last category is where the real shift is happening.
What agents do in production today
In businesses that have moved past the pilot stage, agents typically do four things. They plan workflows — breaking a task like "process this invoice" into a sequence of steps. They access data from your existing systems. They trigger actions — updating records, sending notifications, routing requests. And when they hit something ambiguous or high-stakes, they escalate to a human.
The best agent deployments are not fully autonomous. They are intelligently automated with humans making the final call on anything that matters.
Where agents are delivering real value right now
The use cases that are working today are not glamorous, but they are high-impact:
Customer support triage. An agent reads incoming tickets, categorises them, pulls up account information, and either resolves the straightforward ones or routes complex ones to the right person with full context. CB Insights' Q4 2025 enterprise survey found customer service is the number one area of AI agent adoption.
Invoice processing. Instead of manually matching invoices to purchase orders, an agent handles the matching, flags discrepancies, and sends only exceptions to a human.
Lead qualification. An agent reviews incoming leads against your ideal customer profile, enriches the data, scores them, and routes the promising ones to sales — before a human touches them.
Internal knowledge search. Instead of staff spending 20 minutes hunting through shared drives, an agent finds the right answer in seconds.
What agents cannot do yet
Here is where I want to be honest, because the hype is real and it is loud.
Agents are poor at anything requiring genuine creativity, nuanced judgement, or deep relationship management. They cannot navigate a sensitive employee conversation. They cannot decide whether a borderline deal is worth the risk. They cannot write your brand story in a way that sounds like you.
Gartner predicts that by 2030, 35% of point-product SaaS tools will be replaced by AI agents — not 50% by year-end, as some claim. Deloitte agrees: the full transformation will take at least five years.
The smart move is not to wait, and it is not to rush. It is to start with the right use case and learn by doing. At LF Labs, that is exactly the approach we take with our clients — finding the high-impact, low-risk starting point and building from there.
The Operator's Toolkit: Where to deploy your first AI agent
If you are thinking about deploying an AI agent in your business, here is a practical framework for choosing where to start.
Best starting points:
Support triage — High volume, repetitive, and the cost of a wrong routing is low. Ideal first use case.
FAQ handling — If your team answers the same ten questions every week, an agent can handle that today.
CRM cleanup — Deduplicating records, enriching contact data, flagging stale deals. Boring but valuable.
Meeting scheduling — Coordinating across calendars and time zones is exactly the kind of multi-step task agents excel at.
Budget reality:
A basic agent deployment using existing platforms can cost $500–$2,000 per month in software. Custom-built agents for complex workflows typically range from $10,000–$50,000 for the initial build, depending on integrations and complexity. The real cost is time — expect four to eight weeks to get a production-ready agent your team trusts.
The human-in-the-loop principle:
Smart businesses keep humans in the decision chain. Let the agent handle data gathering, routing, and drafting. Let humans handle approvals, exceptions, and anything requiring empathy. Microsoft reports that over 80% of Fortune 500 companies using AI agents maintain human oversight for critical decisions.
Red flags that mean you are not ready yet:
- Your data is scattered across disconnected systems with no API access
- You do not have a documented process for the task you want to automate
- Your team has not been brought along on why this matters
- You are trying to automate a process you have not fixed yet — AI will not rescue a broken workflow
The Radar: Three things worth knowing this week
1. Voice AI is the fastest-growing segment of generative AI. Funding surged eightfold in 2024 to $2.1 billion, with the market projected to reach $81.6 billion by 2032. Meta acquired voice AI startup PlayAI in mid-2025 and hired leadership from Sesame AI — signalling where the giants think the next interface is heading.
2. Enterprises are consolidating fragmented systems. The 2026 Workflow Automation Outlook from Deloitte and ServiceNow identifies "AI-ready architecture" as the top trend — organisations replacing patchwork software with adaptive foundations for AI decision-making. If your tech stack feels like a jigsaw puzzle, you are not alone.
3. AI agent observability is becoming a category. Fiddler AI raised $30 million in January 2026 to build what they call a "control plane" for enterprise AI — tools that monitor how agents behave, what decisions they make, and where they go wrong. As agents do more, watching what they do becomes its own industry.
From the Field
Here is my question for you this week: If you could hand one repetitive business process to an AI agent tomorrow, which one would it be?
Hit reply and tell me. I read every response — and the best answers might show up in a future edition.
Until next time,
Francois