The Next Chapter of Every Consulting

Today, the Every consulting practice is announcing specialized playbooks for tech and finance companies to go from AI-curious to AI-native. Our consulting team has worked with hedge funds and investors with combined assets under management of over $100 billion, and has trained teams at top tech companies using a methodology that’s earned more than 7,000 stars on GitHub. Below, Every’s head of consulting Natalia Quintero shares what we’ve learned working with these companies—and how any team can get started.—Kate Lee

Was this newsletter forwarded to you? Sign up to get it in your inbox.

---

For the past year and a half, we’ve been doing AI consulting for companies like the New York Times, the hedge fund Walleye Capital, and mental health tech company Headway. What we’ve learned has reshaped how we think about AI adoption—and what it takes to get results.

When we started, we noticed that something fundamental was missing in how professionals were using AI. These tools had drastically improved our own productivity across the editorial, product, and consulting teams, from synthesizing notes and automating meeting actions to extracting value from messy data. But the companies reaching out to us were at a loss for where to get started, or whether AI would even be worth the effort.

So we decided to share what we’d learned from using AI every day. We took on a small group of clients in finance, media, and tech to help them implement AI in their workflows. A year later, we’ve spoken to over 100 companies about their needs and frustrations, and have worked closely with nearly two dozen organizations.

The practices we’ve been teaching have changed as fast as the tech. A year ago, our training focused on prompt engineering, engineering inside of the ChatGPT user interface, and developing robust Projects that referenced up to 20 documents. Now, while those principles and features are still important, they feel ancient.

Today, we’re building custom plugins that connect AI to proprietary data, teaching teams to use Claude Code for end-to-end automation, and deploying agents that run entire workflows without human intervention. The technology has advanced rapidly, and we’ve been developing frameworks to match: compound engineering, which has been recognized by the creator of Claude Code, and agent-native architecture, our guide to building products in this new era.

Our team of applied AI engineers, designers, analysts, writers, and editors are living this future every day. And our experiences have confirmed our long-held theses: We’re moving rapidly to an allocation economy, where individuals won’t be judged by the limits of subject matter expertise, but instead on how well they can allocate and manage AI resources to get work done. The key skills needed to get the most out of AI are the same skills good managers possess—goal setting, clear communication, effective feedback, and constant learning.

Today, we’re unveiling the next chapter of Every Consulting, and to mark this, we’re sharing how we see the state of AI adoption at companies today.

The four levels of AI maturity

Many assume that successful AI implementation requires jumping straight from zero to full automation. Even partial automation can deliver incredible productivity gains.

I grade our clients on a spectrum of automatability—how much benefit a company can get from AI. This doesn’t mean replacing jobs. More often, it entails adding capacity to existing employees by automating well-defined tasks and repetitive workflows using documents and data. And as we have more easily-defined tasks, repetitive workflows, and data handy, they move up the learning curve.

Here is how we see the different maturities of AI usage across organizations:

Level 1: ChatGPT for tasks

You’re using ChatGPT or Claude mainly in chat. You summarize things, write drafts, or generate ideas. You use it almost as a Google Search replacement. (This is a great start, but if this is you, you’re just scratching the surface. Often people stuck at level 1 don’t realize the tasks that they could automate. It’s hard to identify problems to be solved when you live with them every day.)

Level 2: Custom agents—AI tailored to workflows—with permissions

You’ve built some custom prompts or GPTs for recurring tasks. You’re still asking them to confirm what they’re doing. You feel like you’re managing a junior intern.

Level 3: Custom agent does a defined job

Your custom agent no longer asks for permissions. It performs complex tasks like research, data analysis, and project management. The agent can do multiple tasks simultaneously.

Level 4: Custom agent is your screen

The custom agent takes over your development environment, and is all you see. Instead of looking at your inbox or Google Docs, you’re working completely from an environment like Claude or Claude Code You leverage AI skills, have a cross-team GitHub repository to which non-technical team members contribute, and do very little of the “work” that you used to do.

Even in just the past few months, it has gotten easier for people with no prior technical experience to use agents, which should accelerate adoption.

Most companies are still at level one, but want help getting to level four. AI usage is a skill, and like with any skill, it takes time. Companies or teams at level four are rare. They embrace experimentation and often operate in small, nimble environments where they are not bound by the compliance that exists in large companies.

It’s also not always obvious which teams to start with in any given company.

No two industries are the same

Each industry concentrates its highest value work in different departments. In tech, this work often sits with the engineers. In a financial firm, it’s the investors. But the best AI implementation plan doesn’t always start there. In one fast-growing hedge fund, we found that enabling the back office— compliance, operations, recruiting, and administrative teams—with AI would provide the biggest gains. The company was scaling rapidly, and back-office bottlenecks were limiting how quickly the front office could grow.

Additionally, every industry has different levels of documentation, which determines what can be automated. Tasks that can be documented—like generating a report from data—are easy to automate. Tasks that can’t—like deciding what makes a good investment—resist it.

Here are a few of the tasks that we have automated across different roles and industries:

Investment professionals: One hedge fund client wrote out its investment philosophy—what they care about and how they evaluate companies—as a structured skill. Using our compound engineering plugin, they now screen companies and analyze earnings transcripts, financial statements, web data, and regulatory filings in parallel, producing a coherent investment report in minutes. The same work previously took a week.

Researchers and analysts: We have helped investors at private equity firms automate creating investment memos. With the right inputs—documented processes and proprietary data—AI turns two weeks of analysis into a few hours. Using Claude Code and custom skills that encode their investment criteria, analysts can now pull data from multiple sources, run preliminary analysis, and generate draft memos automatically.

Content teams: We built one of the largest media companies in the world a Claude Skill a tool that captures their brand voice, taking rough drafts to editor-approved copy in minutes.

Product teams: Product teams want to automate sorting through user feedback, spotting feature requests, and organizing scattered meeting notes. Plus, when they can quickly vibe code throwaway prototypes to “show rather than tell” what needs to be built, engineering teams have an easier time building and shipping products. With Claude Code and compound engineering, even non-technical product managers can build working prototypes and automate the grunt work of feedback synthesis.

Most importantly, success looks different for each company. For example, our work with a 70-person recruiting firm has saved recruiters five to 10 hours per person, per week. Meanwhile, an investment firm is saving 50 hours per investment memo. In each case, effective AI implementation solves a painful, time-consuming problem.

Documenting the tedious, time-consuming tasks that are also high-value is the best way to start.

The only consultancy working with finance and tech firms

We’ve chosen to focus exclusively on finance and technology firms—industries where the data is rich, the stakes are high, and the potential return on investment from AI implementation is enormous. Our finance vertical is led by Brooker Belcourt, who built the finance arm at Perplexity and previously founded and sold an investment tech company after starting his career at Goldman Sachs, Coatue, and Citadel. If that’s you, we’d love to talk.

In the meantime, here’s my recommendation for any team that wants to get started with AI on their own. Write down a list of your daily tasks, and note which ones you’d assign to a smart intern. Use Monologue to write a detailed job description for how you’d delegate this task. If you’re stuck, ask an LLM to interview you.

This is your master prompt.

To get to level one, paste this prompt into ChatGPT. To get to level two, add additional examples and make it a GPT. To get to level three, turn it into a skill in Claude. And to get to level four: Run it as a scheduled task on Claude Code.

If you find yourself wanting some help, reach out to us. We have a four-step process:

1. Set a strategy: We survey where you’re at: your AI baseline.
2. Build workflows: We build tools that automate the parts of your business that are currently within reach.
3. Train teams: We help you build an AI workforce.
4. Support: Afterwards, we stick around as the chief AI officer.

Whether or not you’re working with us, we’re excited to hear about how you’re using AI. And if you need guidance, we’re just an email away.

---

Want to learn more? Join our consulting information session on February 13. Or if you’re in finance, join our March 13 workshop to learn how we use Claude Code to automate earnings previews, reviews, valuations, and more.

Every is accepting a limited number of consulting engagements for 2026. If you’re interested in working with us, get in touch.

---

Natalia Quintero is the head of consulting at Every. You can follow her on X at @NataliaZarina and on LinkedIn.

To read more essays like this, subscribe to Every, and follow us on X at @every and on LinkedIn.

We build AI tools for readers like you. Write brilliantly with Spiral. Organize files automatically with Sparkle. Deliver yourself from email with Cora. Dictate effortlessly with Monologue.

We also do AI training, adoption, and innovation for companies. Work with us to bring AI into your organization.

Get paid for sharing Every with your friends. Join our referral program.

For sponsorship opportunities, reach out to [email protected].

Help us scale the only subscription you need to stay at the edge of AI. Explore open roles at Every.