What does a Fractional GTM Engineer do?

I build and optimize the AI systems that power your go-to-market — automated outbound, enriched pipelines, smarter CRM, and support automation. I focus on sales, marketing, and support teams because that's where most companies have the most repetitive work that AI can take over. I write the playbooks so your team runs it without me.

What AI consulting services do you offer?

AI Sales Development (lead enrichment, scoring, drafted outreach on arrival), AI Customer Support (automated triage, knowledge base management), AI Sales Enablement (meeting prep, post-call analysis), CRM & Data Enrichment (structured data extraction, deal monitoring), AI Content & Demand Generation (content workflows, outbound sequences), and AI Training for Non-Engineers (hands-on Claude Code training for PMs, marketers, and ops).

How much does AI consulting cost?

The AI Workshop is $5,000 for a one-day intensive. You get a full audit of your current AI usage, a list of quick wins, and a 90-day roadmap. Ongoing Advisory is $8,000/month with a 3-month minimum, then month-to-month. That covers strategy, team coaching, workflow prototyping, and async Slack/email support.

How long does it take to see results from AI consulting?

Quick wins usually land within the first 30 days. Measurable results across your sales, support, or marketing workflows typically show up by day 90. How fast depends on your team's readiness and which workflows we go after first.

Do you build the AI systems yourself?

I design the systems and write the playbooks. Your team builds. I'll prototype workflows and stick around as a fractional advisor to fix what breaks and adjust what's slow, but the goal is always to hand off ownership to your people.

What kinds of companies do you work with?

B2B companies with sales, marketing, and support teams that are ready to adopt AI. My background is 15+ years at Tesla, Mapbox, and HubSpot across growth, operations, and product management. The common thread is teams losing time to manual work that AI can handle.

The Rise of Agent-Centered Design

Every company that builds software asks the same foundational question: who are we building this for?

For thirty years, the answer was obvious. Humans. Design for the human. Understand the human. Iterate until the human can use the thing without getting frustrated.

That's changing. Most teams can see it coming. They just haven't reorganized around it yet.

Why this changes everything

The internet has a new kind of user. AI agents are intelligent enough to engage with platforms and services the way expert humans do — reading documentation, calling APIs, writing scripts, completing tasks. They understand context. They figure things out.

But unlike humans, you can spin up ten thousand of them tonight and shut them all down by morning. (Letting them go is easier too — no HR involved.)

This is what makes agents different from every prior expansion of the internet's user base. Agents are a fundamentally different kind of user — one that scales instantly, operates continuously, and doesn't need a graphical interface to get work done.

The data backs this up. I wrote earlier about what happens when agents outnumber humans on the internet — the numbers were striking then and they've accelerated. Agentic AI reached 35% enterprise adoption in just two years, faster than generative AI (3 years) and far faster than traditional AI (8 years), with another 44% of organizations planning to deploy soon, according to a November 2025 joint report by BCG and MIT Sloan Management Review. McKinsey estimates that by 2030, AI agents will autonomously orchestrate $3–5 trillion in global commerce. Bain projects agents will drive 15–25% of total US online retail sales by 2030. BCC Research puts the AI agents market at $8B in 2025, growing to $48.3B by 2030 at a 43% CAGR. CB Insights found that agent mentions on earnings calls are already up 10x since 2023.

A healthy platform might grow its human user base 10–25% per year. The agentic user base is likely to grow 200–500x over the next five years. The agent-addressable market went from near zero two years ago to trillions projected by 2030. The vast majority of product roadmaps are still centered around designing for the human. That made sense when humans were the only users worth designing for. It makes less sense every month.

AgentX — the newest design era

User Experience (UX) — built for humans. IDEO popularized human-centered design in the 1990s and it became the dominant philosophy for building internet products. UX researchers interview users, run usability tests, prototype in Figma, and iterate until humans can navigate without friction. Making software usable for humans is genuinely hard — people need intuitive interfaces, clear navigation, and hand-holding through flows. At most mature companies, this work takes up 80% of the product roadmap. It's essential, expensive, and slow. It also powered decades of growth, because humans were the only users.

Developer Experience (DevEx) — built for builders. Some platforms realized their most valuable users weren't humans clicking through dashboards — they were developers building on top. Stripe, Twilio, and AWS made developer experience a first-class discipline: clean REST APIs, well-structured HTML documentation, SDKs in every major language, and a dedicated developer relations team. A strong DevEx player could unlock an ecosystem no UX investment could match. Most mature platforms today run a hybrid — heavy UX for humans, a DevEx layer bolted on for builders.

Agent Experience (AgentX) — built for agents. This is the era we're entering, and most product teams don't have a name for it yet. Agents don't navigate UIs and they're not just calling APIs the way developers do — they operate as autonomous users with their own context, goals, and decision-making. AgentX means designing your platform so an intelligent agent can discover what your product does, authenticate securely, execute tasks reliably, and build repeatable workflows, all without a human in the loop. Teams already doing this are shipping MCP servers so agents can authenticate and operate natively, publishing llms.txt files so AI models can ingest their documentation cleanly, packaging common workflows as named skills that agents can discover on directories like skills.sh, and building CLIs that coding agents like Claude Code can drive directly from the terminal. This is the design era where the biggest growth opportunities are sitting right now — largely unclaimed.

The discipline: Agent-Centered Design

I'm calling the practice of building for AgentX Agent-Centered Design. The same rigor that went into human-centered design, applied to intelligent agents — minus the Figma files and the clunky multi-screen UI flows nobody wanted to build anyway.

The design surface is a completely different shape. Make your platform legible via documentation, accessible via API, and operable via standard protocols. That's a smaller engineering lift than building and maintaining a full UI stack, and it ships faster.

Give an agent access to your platform and it will figure out how to use it — no tutorial, no onboarding flow, no help center article. This is both the opportunity and the requirement: agents reward platforms that are well-structured and punish ones that aren't. There's no human to forgive a confusing experience.

What Agent-Centered Design actually requires:

Markdown-first documentation. Agents understand text and structure. They don't understand animated docs portals. Your documentation needs to exist in clean Markdown, accessible via a /docs endpoint or an llms.txt file agents can ingest directly.

MCP support. The Model Context Protocol is the emerging standard for how agents authenticate and operate within external platforms. MCP server downloads went from roughly 100,000 in November 2024 to over 8 million by April 2025. Stripe was early. Block, Bloomberg, and hundreds of Fortune 500s have deployed it. If you don't have an MCP server, you're invisible to a fast-growing class of agentic workflows.

Skills and packaged templates. Pre-packaged recipes that define how to accomplish common tasks on your platform. Directories like skills.sh from Vercel are emerging as the distribution layer — an app store for agent capabilities.

A native CLI. Coding agents like Claude Code and OpenAI Codex operate in terminals. A well-designed CLI lets an agent drive your platform, query it, update it, orchestrate it, without any human in the loop.

Full API coverage. Not just APIs for what you've historically exposed. Full coverage of everything your platform can do. If it's only available in the UI, it's inaccessible to agents.

Who's already doing it

Ahrefs spent years building a best-in-class product for human users. Great UI. Deep data. Relatively expensive subscription. And until this week: minimal API access.

That's a problem when agents want to run SEO queries. They can't click around a dashboard. They need to call an endpoint.

Ahrefs just launched their first developer API for paid subscribers. They're not abandoning their human users — they're opening a second growth surface that didn't exist before, one that will compound at a rate their seat growth never will.

AgentMail goes further — they skipped the human user entirely. It's an email platform built exclusively for AI agents. Agents get real email addresses, spin up new inboxes via API, and communicate autonomously without any human interface involved. Their tagline says it plainly: "It's not AI for your email. It's email for your AI." They raised $6M from Y Combinator in March 2026.

This is Agent-Centered Design at full expression: a product where the agent isn't an edge case or a power user. It's the only user.

Start now

Audit your API coverage. Ship an MCP server. Publish an llms.txt. Build a CLI. Package your capabilities as skills and get them onto directories where agents can find them.

Companies that went mobile-first in 2011 didn't see the full payoff until 2014–2015. The ones who waited until 2015 lost years of compounding they never got back. The same dynamic is playing out now, just faster.

If you're focused purely on human user growth right now, you're squeezing a stone. The sponge is sitting right next to it.

Start thinking about how agents can utilize your platform. What would a consumption-based model look like if agents were your fastest-growing user segment? That's the question that fuels the next decade of growth — and the teams asking it now are the ones who'll look obvious in hindsight.

Colin Gillingham is a Group Product Manager at HubSpot and the builder behind PhoneScreen.ai, TractorBeam, and UXLoops.