The Huddle: Issue 05

A biweekly look at what's changing in AI, ways to apply it today, resources you didn't know you had, and sharing wins.

Unpacking Agents

It’s Friday and time for another Huddle. This one is a big one for us, as today is the one year anniversary of Trailblaze Labs. I’ll save the mushy speech, but we’re grateful for the companies (and people behind them) who trust us to help translate the most complicated era in the history of technology. It’s a joy, a challenge, and we are thrilled about what’s around the corner - which brings us to AGENTS.

This issue is everything I'd want a business leader to know about agents right now. What just happened, what an agent actually is (and isn't), and five real ones we've built with clients in the last few months. Buckle your chinstrap.

Quick note: I'm going to use the word "agent" a lot. By the time you finish this issue, you'll know exactly what I mean by it, promise.

VOCAB CORNER: AGENT VS. CHATBOT

If you're going to navigate the next 18 months of AI conversations, the chatbot-vs-agent distinction is the one to internalize.

A chatbot answers your questions. It waits for prompts, doesn't carry memory between sessions, and lives inside one window. An agent does the work. It has access to your tools and systems, holds memory across runs, can be triggered by a schedule or a Slack message, and can take real actions like sending an email, updating a spreadsheet, or pulling data from your CRM. A chatbot answers. An agent acts. A chatbot lives in a window and waits for you to open it. An agent has its own logins, its own memory, and its own reasons to be working at 6am before you've had your first coffee.

HEADLINES THAT ACTUALLY MATTER

OpenAI's Workspace Agents (April 22)

OpenAI introduced Workspace Agents inside ChatGPT on April 22, framing them as the direct successor to Custom GPTs. They're built on Codex, run in the cloud, and plug into Slack, Salesforce, Google Drive, Notion, and Microsoft 365. You build one, share it with your team, and it keeps working even after you've closed the laptop. Free during research preview through May 6, then a credit-based pricing model takes over. Available on ChatGPT Business, Enterprise, Edu, and Teachers plans. The example OpenAI led with was their own sales team using an agent to pull call notes, qualify leads, and draft follow-ups directly inside the rep's inbox.

Why this matters: Custom GPTs were one-person creations that waited for someone to ask them a question. Workspace Agents are shared, persistent, and able to take action across multiple tools. That's the chatbot-to-coworker change, and it's available to your team this week.

Google's Gemini Enterprise Agent Platform (also April 22)

On the same day, Google announced the Gemini Enterprise Agent Platform from the Cloud Next stage in Las Vegas. It's the evolution of Vertex AI, designed as a single place to build, govern, scale, and optimize agents across an organization. They've baked in 200+ models in the Model Garden (including Anthropic's Claude), a low-code builder called Agent Studio, an Agent Registry to track every agent in the company, and an Agent Gateway that enforces security and prompt-injection protections. Box, Workday, Salesforce, and ServiceNow are all building partner agents on top of it.

Why this matters: Google is making the case that if you're going to deploy agents at any kind of scale, you'll want them on a platform that handles identity, security, and governance from day one rather than something you stitch together later.

Microsoft Agent 365 (live today, May 1)

Microsoft's announcement is less about a new agent you'll go build and more about the wiring underneath. Agent 365 is the layer that lets IT and security teams see every agent running across the business, who built it, what it can touch, and whether it's behaving. If your team has been quietly experimenting with agents and your IT lead has been quietly losing sleep over it, this is the answer they've been waiting for. It's also a signal: Microsoft is betting that the bottleneck for agents inside bigger businesses isn't going to be capability, it's going to be trust. Their internal team is already running 500,000+ agents on it, with the heaviest use in research, sales intel, customer triage, and HR.

The other Microsoft thread worth tracking: TechCrunch reported on April 13 that Microsoft is quietly building an OpenClaw-style agent for 365 Copilot, expected to debut at Build in June. Which brings us to the curveball.

The "OpenClaw" Curveball

If you've spent any time on AI Twitter recently and seen people posting lobster emojis, that's OpenClaw. It's an open-source AI agent you run on your own hardware (most commonly a Mac Mini sitting in someone's home office or data closet) and connect to the messaging apps you already use, like WhatsApp, Telegram, and iMessage. It launched in late 2025, hit 214,000 GitHub stars by February, and inspired thousands of developers to set up dedicated Mac Minis as personal "Chiefs of Staff." Its creator, Peter Steinberger, joined OpenAI in February to lead their personal agents work.

Why this matters: while OpenAI, Google, and Microsoft are racing to be your enterprise agent platform, OpenClaw is showing there's a parallel appetite for "I want my own agent that lives on my hardware, knows my life, and answers my texts." For most businesses, the enterprise platforms are still the right answer (and we'd recommend starting there). But your ambitious team members may be testing their own “Claws” and it’s important you have tight policies around that to protect enterprise data.

FIVE DIGITAL TEAM MEMBERS WE'VE ALREADY BUILT

We're already deep into agent work with clients. Here are five real ones, currently doing real work. We’ll keep names/companies a bit “blind” but I think you’ll get the gist.

The Data Analyst. A client was paying for PowerBI plus a data analyst to refresh dashboards weekly, monthly, and quarterly. We built an agent connected to their data warehouse that takes natural-language questions ("show me Q1 revenue by product line, year over year") and responds with the chart and the narrative behind it. The agent doesn't just visualize. It explains what changed and where to look next. This agent even drafts weekly communication to their clients explaining anomalies and tracking trends.

The HR Generalist. A client whose HR lead handles benefits, policy, and people questions but doesn't have a legal background. The agent has read the entire employee handbook, the relevant state and federal regulations, benefits administration, and the company's internal policies. When a question comes up about FMLA, ADA accommodations, or how to write up a sensitive conversation, she has a thoughtful first draft in 30 seconds with sources cited. She still makes the call. The agent makes sure she's not making it cold.

The Leadership Coach. We trained an agent on the personality profiles of an entire organization (yes, 200+ people). We’re talking about the org chart, the team dynamics, and the founder's leadership philosophy. Any leader on that team can now ask it for help prepping for a 1:1, a quarterly review, or a hard conversation. The advice is tailored to who they're actually talking to, not generic best practices from a Harvard Business Review article. They’re using it to make hiring decisions, vet candidates for internal promotion, and shape teams intentionally.

The Chief of Finance. A client with a Director of Finance retiring after 20+ years was facing a six-month gap before a new hire could fully ramp up. We built an agent loaded with their accounting practices, vendor relationships, the why-we-do-it-this-way context, and the policy quirks that live in someone's head and not in any manual. The new Director will inherit a teammate who already knows the playbook.

The Private Eye. A competitive intelligence agent built for a brand in a crowded category. It tracks competitor product launches, industry events, pricing changes, and key hires. It runs on a schedule and pings the marketing leader via email the moment something material happens. No more learning about a competitor's launch from a LinkedIn post three weeks late.

What catches clients off guard every time we kick one of these off is how buildable they are right now. We're talking days or weeks, not quarters. The barrier isn't technology, it's clarity on what you'd actually want a digital teammate to do for you. Once you have that, we can usually have a working version in someone's hands within a sprint.

TRAINING TIP: PICK THE BORING ONE FIRST

The mistake I see businesses make is building the cool agent before the useful one. The cool agent is flashy ("an agent that runs my whole company"). The useful agent is unglamorous ("an agent that handles the Monday morning report I hate doing").

When clients ask where to start, I tell them to find one workflow that hits three conditions: it happens repeatedly, it currently lives across two or three different tools, and the person doing it doesn't enjoy it. That's the agent that should exist. Build that one first, use it for two weeks, then build the next one. The momentum compounds faster than you'd expect, and the boring one is what teaches you what you actually want from your second, third, and fourth.

YOUR CHALLENGE THIS WEEK

What's on your agent wish list? If you could hire three digital team members in the next quarter, who would they be and what would they do? Reply to this email with even half-formed ideas. I'll send back what we'd actually build first, what we'd push back on, and where we've already built something close for someone in a similar spot. This is genuinely the most fun part of the job, and we're not the only ones who think so (some clients now keep a running list).

THE BUZZER

Agents in 2026 remind me of a young, talented team in training camp. The plays are starting to work and the chemistry is showing up, but you'll see a missed handoff every once in a while. That's expected. The teams who draft early and develop their own talent are going to look very different in two years than the teams who sit on the sidelines waiting for a finished product.

We're in early innings here (yes I’m mixing sports metaphors, sue me), and the work we're doing with clients today is going to compound into a serious advantage by next spring. If you've got an agent in mind, even a half-formed one, reply and tell us. We’ll tell you whether it's a Day 1 build or a Year 3 build, and if it's the former, we'll help you get started.

Thanks for huddling with us to start the weekend,

Bryce Stuckenschneider

Founder, Trailblaze Labs

ABOUT TRAILBLAZE LABS

We started Trailblaze Labs because we saw a gap forming between what AI could do and what most businesses knew how to do with it. Turns out, that gap is where we live now. We help business leaders set real AI strategy, train their teams to use these tools with confidence, and build the workflows that actually move the needle. Trailblaze translates AI to the language of your business.

If something in this newsletter sparked a question, gave you an idea, or made you want to push back on something I said, I want to hear about it. Seriously. Reply to this email, connect with me on LinkedIn, or just say hi. I read everything and I love a good friendly argument about where this is all headed.

Let's keep building.