Coordinating project work across a larger team takes time and attention. Jared Threw built a human-AI collaboration model at House of Growth that gives people and agents a shared process.
House of Growth is a Denver-based performance marketing agency led by founder and CEO Jared Threw.
Much of its project work used to be done by hand and shared across a larger team of part-time contractors. Threw wanted the work to run on a more consistent system.
Today, a lean team of eight runs the agency’s client work alongside 20 active agents. In Threw’s human-AI collaboration model, agents handle most of the grunt work (think first drafts and research), while people lead strategy and the final review.
We caught up with Jared Threw to hear how he connected agents to the agency’s real tools, and why every piece of their work still passes a human check.
How Jared Threw Built Human-AI Collaboration Into His Agency’s Delivery Model
Summarize this article with AI ClickUp Brain not only saves you precious time by instantly summarizing articles, it also leverages AI to connect your tasks, docs, people, and more, streamlining your workflow like never before.
Summarize this article for me please
Why Did House of Growth Rethink How Project Work Gets Done?
The contractor approach kept work moving, but it also left Threw coordinating a larger group without a reliable, consistent process. In his own words:
Most of this work was done manually and with a larger team of part time contractors. This was hard to manage though and felt less in my control. It also felt less systematic and thus, less consistent.
Jared Threw, Founder & CEO at House of Growth
What Threw describes here is a coordination problem: keeping each project consistent depends on someone tracking every step. He decided that human time should instead be going to the steps where judgment changes the outcome.
Summarize this article with AI ClickUp Brain not only saves you precious time by instantly summarizing articles, it also leverages AI to connect your tasks, docs, people, and more, streamlining your workflow like never before.
Summarize this article for me please
Which Work Stays Human at House of Growth?
As a first step, Threw put the process in writing.
He developed SOPs that define the human and AI steps for every area of the business, from campaigns to sales to operations.
He then built agents to carry out those steps, close to 30 so far, each with a clear purpose. They work in pods for SEO and GEO, sales, campaign development, and management.
One small team sits around those pods: three ad managers, an SEO and GEO specialist, an analytics and attribution specialist, a GTM engineer, a web developer, and Threw as CEO and senior campaign manager. He explains how the work is divided between them:
We now leverage an AI + human hybrid model. Humans are the senior strategists and final reviewers. Most of the middle part of the processes to complete projects are completed by AI. This has allowed my lean team to operate at a level not possible before.
Jared Threw, Founder & CEO at House of Growth
For Threw’s team members, that means spending less time producing each piece by hand and more time setting the brief, reviewing what agents produce, and deciding what is ready for the client.
The split required Threw to build a solid operating layer around the agents.
Summarize this article with AI ClickUp Brain not only saves you precious time by instantly summarizing articles, it also leverages AI to connect your tasks, docs, people, and more, streamlining your workflow like never before.
Summarize this article for me please
Why House of Growth Wrote the Process Down Before Building Agents
An agent follows whatever process it’s given. If Threw had automated the old way of working, his AI agents would have carried over the same inconsistency he was trying to fix.
With the process written down first, every agent started from one shared standard. Building on that standard took five steps:
Replace fixed automations with agents
The starting point was the agency’s automation, which ran on if/then rules that follow the same path every time. Almost all of that has moved to agents that can read the situation and adjust their steps accordingly.
A lead form shows the difference. When a prospect submits one, the lead lands in the CRM space, and the Lead Enrichment Researcher automatically researches the company and the person, so Threw has context before a follow-up or a meeting.
As you’d expect, we leverage ClickUp’s AI features extensively, almost completely removing the use of the more manual if/then automation flows in favor of more dynamic AI agents.
Jared Threw, Founder & CEO at House of Growth
Write agent instructions by talking through the workflow out loud
Each agent then needs instructions, and almost every one at the agency is built from scratch.
The first draft is built from a transcribed snippet that helps to preserve all the nuance. Threw talks through a process out loud with a voice-to-text tool, such as the one in Brain MAX or Wispr.
Time is precious for me, so I typically use a voice to text annotation tool like the one within Brain MAX or Wispr. This allows me to talk/process out loud and ramble a bit, getting everything out of my brain far faster than if I typed it all out.
Jared Threw, Founder & CEO at House of Growth
Skills shape how the agents sound. The Sales Outreach Specialist and the Reddit agent write in Threw’s style, and the Marketing Copywriter uses skills for the brand voice and for scripts that sound like him.
Instructions only go so far without access to the real work. The majority of the agency’s software stack connects to its agents through native integrations or custom MCP server connections.
SEO agents run audits in SE Ranking
Content and GEO Specialist creates and optimizes pages in Frase
GTM Engineer reviews outreach campaigns in Clay and Instantly
Ads Agent pulls from the ad platforms, Google Analytics, and Search Console together
Native connections, ClickUp’s Integrations in this case, fill in the rest of the context:
Between these MCP connections, and the ClickUp native features, like Gmail access, Google Drive access, Google Calendar access, AI notes/meeting transcripts, chat conversations, task activity, and more, the agents became as capable as any human employee.
Jared Threw, Founder & CEO at House of Growth
The connections have limits, and the team works around them. Agents can’t build tables in Clay directly yet, so the Clay Audience Builder writes a detailed plan for each table, making it much faster for a team member to build.
Every client has a master context doc, and the agents read it before they work. For the copywriter agent, that step is part of its instructions:
It is also instructed to review a client’s master doc before completing a client copywriting assignment so that it can adjust the writing style to fit each client’s unique style and guidelines.
Jared Threw, Founder & CEO at House of Growth
The campaign strategist uses the same doc to see what has been tried and what worked. The Operations and Knowledge Manager keeps those docs, the agency’s SOPs, and its internal master doc current, using meeting notes, chats, task comments, and emails.
Organize agents into pods with a lead
The last step links the agents together.
They work in pods for SEO and GEO, sales, campaign development, and management, and the SEO and campaign pods each have a lead orchestrator who splits the work and assigns it.
For example, we have a pod of agents that specialize in different areas of SEO, with a SEO strategy expert being the orchestrator that assigns to the correct SEO specialist within the pod.
Jared Threw, Founder & CEO at House of Growth
The campaign pod runs the same way. The Omnichannel Campaign Strategist develops a cross-channel strategy and then creates subtasks for the copywriting, design, and ad agents. Social content adds a human gate in the middle: the Social Ideation and Strategy Engine writes post briefs, and it creates subtasks for the visual and the copy only after Threw approves the output.
See another agent roster at work: A partner services lead walks through six custom agents for renewal forecasting, scoping and pricing, approvals, project setup, weekly reporting, and meeting notes. Together, they save seven hours a week.
Summarize this article with AI ClickUp Brain not only saves you precious time by instantly summarizing articles, it also leverages AI to connect your tasks, docs, people, and more, streamlining your workflow like never before.
Summarize this article for me please
How Does House of Growth Turn Team Feedback Into Better Agent Instructions?
An agent’s first version is a starting point.
At House of Growth, each agent is built quickly and then put to work on real client projects. The team’s reactions to its output shape what the agent does next. That feedback process takes place in the same comments and conversations where the team already discusses the work.
When a teammate flags a problem, the conversation goes back into the agent’s setup, so the fix lands in its instructions, tools, and skills. Threw explains the loop:
When I refine my agents, I often will tag “Brain” in a comment or conversation where feedback has been provided. This has made it easy for my team to provide feedback on the AI outputs the agents are producing, and then have ClickUp’s AI Brain review the feedback and implement the necessary changes to the agent’s instructions, tool access, skill access, etc, to refine it for future work.
Jared Threw, Founder & CEO at House of Growth
Because of that loop, the agent system keeps improving after each run, making the same problems less likely to come back.
Summarize this article with AI ClickUp Brain not only saves you precious time by instantly summarizing articles, it also leverages AI to connect your tasks, docs, people, and more, streamlining your workflow like never before.
Summarize this article for me please
Why Every Agent Output Still Gets a Human Check
Every agent at the agency, from the SEO pod to the sales agents, gets the same checkpoint.
For all of these agents, I have implemented a human QC step before any AI generated work is published. This ensures quality and avoids misguided AI content from being auto published without proper oversight.
Jared Threw, Founder & CEO at House of Growth
That checkpoint is where the team’s expertise does its most visible work. A reviewer can spot when an output misses the brief, misreads campaign data, or calls for a different strategic choice, and send it back before a client sees it.
Summarize this article with AI ClickUp Brain not only saves you precious time by instantly summarizing articles, it also leverages AI to connect your tasks, docs, people, and more, streamlining your workflow like never before.
Summarize this article for me please
What Changed for House of Growth’s Clients?
The clearest test for Threw’s agents came with two new boutique hotel clients that needed SEO help.
Once both clients were set up in the agency’s SEO tools, the SEO Strategy Lead agent audited each site and scoped subtasks for the specialist agents in its pod. In less than one business day, the pod had finished the audits and produced content across SEO and GEO.
The team’s review came next, and that is where the timeline changed:
Paired with the human review phase, we were able to deliver a month worth of work in less than a week, and because of the MCP server connections to our SEO tools, the outputs were accurate and data-backed.
Jared Threw, Founder & CEO at House of Growth
Paid search produced a second example. A boutique hotel client’s Google Ads campaigns had started to perform poorly, and the manual audit was slow because the patterns were spread across several tools.
Leveraging our Ads agent, it was able to audit the Google Ad account, each campaign, Google Analytics, and the internet for market variables. It identified patterns and opportunities that we had missed and helped orchestrate the creation of new campaigns and new copy leveraging our copywriting agent.
Jared Threw, Founder & CEO at House of Growth
By his account, the faster recovery saved the client thousands of dollars in lost revenue.
How House of Growth’s delivery changed: before and after agents
Workflow
Before
After
Ad campaign planning
Audits, market research, strategy, keywords, copy, and creative took the ad team about a week.
An agent pod does that work in a couple of hours, and the team spends the time saved on review.
New-client SEO
Audits, scoping, and content were done mostly by hand.
Agents finish audits and drafts in under a business day. With human review, a month of work ships in less than a week.
Campaign diagnosis
The team audited weak campaigns manually and compared data across tools.
The Ads agent checks the ad account, Google Analytics, and market factors together to find the cause.
Team capacity
More client work meant more people and more manual steps to coordinate.
Each team member can take on more clients without a drop in quality.
Summarize this article with AI ClickUp Brain not only saves you precious time by instantly summarizing articles, it also leverages AI to connect your tasks, docs, people, and more, streamlining your workflow like never before.
Summarize this article for me please
Why Does House of Growth’s Model Matter Beyond One Agency?
In the 2026 AgencyAnalytics benchmarks report, 38% of agencies used agentic AI for workflow automation, and 58% had increased human review of AI-generated work. House of Growth sits in both groups.
The open question is what that review costs.
Optimizely found that 76% of marketing leaders spend at least three hours a week editing, fact-checking, or correcting AI output. Those hours fix one output at a time unless the correction reaches the system that produced it. At House of Growth, the feedback loop sends each correction back into the agent’s instructions for the next project, and MCP connections keep the data accurate.
The structure behind Threw’s model matches what researchers now recommend.
The Conference Board’s September 2026 framework advises breaking work down task by task, deciding what belongs with AI, with people working alongside AI, or with people alone, and naming one person accountable for each workflow. House of Growth’s SOPs handle the first part, and Threw’s team, as senior strategists and final reviewers, covers the second.
The review step also puts the team’s time where workers say human skills matter most.
Summarize this article with AI ClickUp Brain not only saves you precious time by instantly summarizing articles, it also leverages AI to connect your tasks, docs, people, and more, streamlining your workflow like never before.
Summarize this article for me please
What Should Agencies Decide Before Building a Human-AI Collaboration Model?
Agencies are at very different stages with AI. Promethean Research’s 2026 benchmark found that 34% of 119 digital agencies had implemented it across the business, and 28% were still implementing it.
A few decisions shape how well the human-AI model works.
Which workflow goes first?
Start with work that repeats often and has an output a reviewer can check against a clear standard, such as a site audit or a campaign report. High-stakes strategy work is tempting because the payoff looks bigger, but it is harder to check and riskier to get wrong in front of a client.
Who owns each agent, and how will you measure the results?
Every agent needs one person responsible for its results and its improvement. Without an owner, feedback piles up in comments, and the instructions stay the same.
Time saved is the obvious measure, and it can be a misleading performance metric on its own. Deloitte suggests tracking error rates and output quality alongside speed. Count the hours spent checking and correcting agent work, so the real gain is visible.
Clients notice when the way work gets made changes. In the 2026 AgencyAnalytics benchmarks, 45% of agencies said they were being more open with clients about how they use AI.
Decide early how you will explain which steps agents handle and where people still review the work.
Where will the freed time go?
Saved hours tend to drift into whatever is urgent unless someone decides where they go. There are three common choices, and each one changes the business differently.
More clients per person. This raises revenue without adding headcount, but it also raises the stakes on review, because each person now checks more agent output across more accounts. House of Growth took this route: Threw says each team member can now take on more clients without a drop in quality
Deeper strategy for existing clients. The time goes back into the accounts you already have, through more analysis, more planning sessions, and closer attention to results. Retention and account growth benefit, though the gain is harder to show on a capacity report
Faster testing. Campaigns can be launched, measured, and adjusted in shorter cycles. Clients get answers sooner and spend less on approaches that don’t work. Threw notes that his team can now test and refine campaigns at a speed that would previously have cost clients far more
Whichever path an agency picks, it helps to prove the gain before counting it. The Conference Board’s 2026 framework advises against treating freed time as a financial gain until the organization can show how that time was used and what measurable result it produced.
Summarize this article with AI ClickUp Brain not only saves you precious time by instantly summarizing articles, it also leverages AI to connect your tasks, docs, people, and more, streamlining your workflow like never before.
Summarize this article for me please
What’s Next for House of Growth?
House of Growth has already answered most of these questions. Threw leads strategy and final review, agents work from live client data, and the freed time goes to more clients per person and faster testing.
The next stage builds on that same operating layer.
His team is exploring how AI skills can refine the agents further. It is also testing ClickUp Artifacts for interactive HTML reports and playbooks, personalized pitch decks built from a template, and client performance dashboards.
We will always include the human element in our work, clients want that, but we are continuously testing and refining how work can look using a human + AI hybrid model.
Jared Threw, Founder & CEO at House of Growth
House of Growth started with a coordination problem and a decision to make the work more systematic. That decision still shapes every new addition: it goes into the written process first, then into the agents, with a person reviewing the result.
Summarize this article with AI ClickUp Brain not only saves you precious time by instantly summarizing articles, it also leverages AI to connect your tasks, docs, people, and more, streamlining your workflow like never before.
Summarize this article for me please
Frequently Asked Questions
What is human-AI collaboration?
Human-AI collaboration is a way of dividing work so AI handles defined execution while people set direction, apply judgment, and own the result. In Microsoft’s 2026 Work Trend Index, 86% of AI users said they treat AI output as a starting point and stay responsible for the thinking. The model works best when the split between human and AI steps is written down and reviewed as the work changes.
What are the benefits of human-AI collaboration?
The main benefits are faster delivery and more time for work that needs judgment. In the 2026 AgencyAnalytics benchmarks, 79% of agencies said AI saves them five or more hours a week. The same agencies increased human review at the same time, which suggests the gains hold when people stay responsible for quality.
What are the risks of human-AI collaboration?
The biggest risks are inaccurate output and work that reaches clients without a proper check. In a 2026 Optimizely study of 2,003 marketing leaders, fact-checking and hallucination review created more extra work than any other factor, and 25% admitted to publishing off-brand AI content under deadline pressure. A required review step before publishing reduces both risks.
What is the difference between human in the loop and human on the loop?
A human in the loop works alongside AI on each task, while a human on the loop supervises work that AI systems plan and carry out. Deloitte describes the shift from one to the other as multi-agent systems take on more execution. Both models still need clear roles, monitoring, and a person accountable for the outcome.
What is an MCP server, and how do AI agents use it?
An MCP server connects an AI agent to an outside tool or data source using the Model Context Protocol, an open standard introduced by Anthropic in 2024. Through that connection, an agent can read data from tools such as an SEO platform, an analytics account, or a CRM, and take actions where it has permission. This lets agents work from live data instead of a blank prompt.
Everything you need to stay organized and get work done.
Summarize this article for me please
