What AI Native Consulting Actually Looks Like: The 4 Stages
With Real AI Agent Examples, Workflows & Templates
You can tell a lot about where people are with AI by the words they use to describe the work they’re doing, how they’re presenting it, or even how they’re delivering it.
And there are probably a lot of “AI transformation” firms that still run on a manual back office and Google Docs.
I started taking ai serious in work in early 2023 to constantly improve the way I work, deliver value, and increase the value I create. From research, to auditing sales team calls, to combing through a pile of information for the needle in the haystack.
A lot of firms right now are pressured to become “AI native” and a lot of people talk about it but I want to peel back the curtain and share a first hand experience of what I’m actually doing in my own firm, and what I’m seeing as I take clients through this. Because it is a journey. It’s not something you can just install in one day.
I think there’s a version of a consultancy or agency that is genuinely worth building.
It gets you out of the trap of “how do I avoid getting replaced?” and helps you create a new kind of firm. One that delivers more value, helps clients go further, increases your margins, and could potentially create a category of one for you so that you can price accordingly.
And the only way to do that is to embrace the tech for good.
Everything I do inside my firm runs on AI (and yes the more I work with it, the more I want to move toward my own local models.)
But before we get into the tech, I want to share a framework of migrating your work to become ai native so you can build the firm of the future from a systems perspective, instead of chasing the latest fancy tool or automation or claude md.
(Most of what’s being posted online, sold through “overnight experts” right now is, honestly, garbage) Just copy/paste what they are saying back to your own AI agents and reverse engineer what they do.
It will save you from inheriting a system that may not work for you or worse downloading and using a skill md that may or may not have malware in it.
Alright, there are 4 stages to migrate your company to becoming ai native and they’re sequential.
And while each stage is a progression, it’s up to you to decide which stage you want to operate at and which one you’re comfortable at.
The 4 stages, in order are
Enhancement
Augmentation
Parallel Operations
Autonomy
Let’s go through each.
Stage 1: Enhancement
This is where everyone should start. And honestly, this is probably where you’ve been. Especially when people say, “I use AI in my work.”
There’s a lot happening in the market right now. People are using AI in their work and either telling their employer about it or not: this gets called “shadow AI” where team members using it to operate and execute without telling their bosses aka you. You may not actually know your team is running this. And it can be potentially dangerous that your company info is being used or uploaded in an unprotected way (but thats for another post)
Enhancement is the simplest stage. You layer AI on top of what you’re already doing to work faster, go a bit deeper, and create real efficiencies.
What doesn’t change here:
the structure of your operations
your offerings
your price points
You’re not redesigning your work. You’re just improving how you’re already doing it.
For example, in the past, when I would audit sales calls from Gong, or review team calls for large clients, I would need to find specific needles in the haystack and cross-reference them to my own frameworks.
Honestly, I could listen to maybe a handful of calls a day.
But listening to the last three or six months of calls with the speed and agility I needed would have been nearly impossible to do in the timeframe I needed to provide insights to the team.
So I would use my own judgment, with my frameworks, and my own way of auditing and analyzing calls with ai to make the work efficient.
You’re probably already doing this right now. Maybe for client prep, maybe for research. Maybe just for cleaning up the CRM.
A lot of the tools you use are becoming “AI native.” They have the little agents or co-pilots embedded inside them. Nothing new. But this is a real time saver. Instead of taking four hours to do something, you’re now taking 20 minutes.
The dark side of this:
you might be doing more 20 minute work.
So instead of doing 1 thing for 4 hours, you might be doing 12 or 15 things for 20 minutes each using up the same amount of time. You may not be really getting time back, but you might be getting more work done. Burnout is real.
Enhancement gives you more capacity, and it’s up to you how to use it.
But this stage has its own ceiling.
Yes, you become more efficient, your margins might improve, and you might do more by the hour.
But it doesn’t fundamentally change the way you run your business as AI-native.
It doesn’t change your operational structure, your margin structure, your team structure, your leadership structure, or your go to market.
Enhancement is part of the foundation. It’s not the actual infrastructure that’s going to get you to the next level.
It’s a prerequisite for augmentation.
Stage 2: Augmentation
This is where things start to change. And this is less about delivering more output and focused around packaging your IP to deliver at scale.
One of the biggest internal lessons for me, and it took me 2 years to create the habit, was to write my thinking every day. To actually have that daily practice. To think about my own thinking every single day, which is just another way of saying “writing:.
It really did take me two years to develop this habit consistently and it finally stuck after doing Seth Godin’s and Ship30’s programs.
The difference between doing more work and scaling your IP comes down to this:
you define your own methodology of how you deliver value to clients.
And if you can compound how AI helps you do that more effectively or go deeper with clients, you’re effectively delivering your intellectual property at scale without needing to be there for every step.
A lot of the work you’re doing right now is based on a method, a protocol, or something that has worked in the past to get results for clients. If you don’t document it, or you don’t know how it’s actually being done, you’re going to a) keep doing it yourself, or b) someone on your team is going to do it, c) your team plus tech is going to do it.
OR option D) you can start leveraging agents to fill in the gaps and help clients go further.
Here we have to look back at what I wrote about in the Services Stack. You can’t train AI to the level of your judgment, your particular insight, or your nuanced takes on client deliverables. But you can train it on simple execution. Things that just need to get done. You can also train it on prerequisite work like template-style strategy so it can fill in the blanks for you.
It can help you go further.
BUT the last 10% comes through your intuitive judgment of what the client actually needs and the strategy and direction you take them in.
There’s nuance here. But once you’re able to deliver what would have taken 20 to 40 hours in half a day, instead of weeks, this stage is very powerful. (I speak from experience when I share those numbers)
For example, I’ve been able to write about my own methodologies. I’ve published several books. Most weren’t on WSJ or NYT, but they helped clients get real revenue, and that’s all that matters to me. (One was an Amazon bestseller. But that’s a different story for a different time.)
Regardless, I trained different ai agent skills around different pieces of my methodology. It’s not just one master skill md. I have a waterfall of nine to twelve different pieces of work I run when I work with clients. I trained the lower mechanics, the parts I know AI can do better than me:
Deep research
Competitive landscaping
Surfacing insights based on the specific questions I ask and the questions I review with clients.
I always know and have a feel for where we’re going to take the client and how we’re going to transform the business bc I’ve trained literally thousands of entrepreneurs and have seen behind the scenes of hundreds of businesses. From my perspective, I know what works and what doesn’t. But I’m leveraging AI as a co-partner to help me co-create things I couldn’t do on my own with this level of speed.
Speed to value is the first thing youll feel, then it’s speed to NET new value.
But first, lets discuss speed to value:
With this type of setup I’m thinking and working at the speed of my thoughts.
But speed is a side effect.
The thing I’m now able to do, if I can get to value with my own intellectual property faster for clients, is start asking: what’s the net new value I can add?
What new things can I start doing or delivering, or helping clients achieve, that I couldn’t do before?
Because all my bandwidth, time, energy, and money was sucked into work that took weeks. And now it can take hours.
This is where you can now create net new value for yourself and your clients. If you want a deep dive on Net New Value - watch this
The major prerequisite for this stage:
your intellectual property has to be codified before you can deploy this stage
If you’re listening to this and you’ve never written down your own thoughts about how you think, that is your number 1 strategy and tactic right now.
Use that organ in your head and start writing what you think, how you process information, and what makes you uniquely useful.
Otherwise, you can’t separate what an AI agent can do from what it can’t. And if you don’t know that difference, you’re going to be in real trouble.
Stage 3: Parallel Operations
What I just described in Stage 2 is an AI agent working with you to deliver. But the other areas of the business actually matter too.
Let’s dissect what a business model actually looks like. I really like using the Business Model Canvas for this.
When you look at a business model, you can see the different criteria that make a model:
your value proposition
how you keep client relationships
how you reach clients
who your ideal client profile is
how you make money
your pricing structure
your key activities
your key vendors and partnerships and resources
and your cost structure (your expenses, your overhead)
I like to codify the business into three parts.
Operations: administration, finances, strategy, team.
Product: client fulfillment, lifetime value, retention, R&D, and inventing new ways to help clients go further.
Marketing and sales: getting clients and keeping clients.
Rinse and repeat.
(I combine marketing and sales together, by the way)
If you’ve codified how you deliver under product, you can start codifying how you operate.
You can codify your key activities, your key customer relationships and how you keep in touch with them. You just need to know where AI falls short and where you need to come in.
The best operators I’m collaborating with, working with, or am friends with right now, the CTOs, leaders in tech firms, and other founders, are using AI with a level of autonomy that runs in parallel operations with a human in the loop.
The agents report to you and work alongside you.
So in Stage 3, you’re running an operational layer that reviews your key activities every day and runs alongside you. The agents execute, direct outcomes, and report back.
You’re literally staffing a small team, or a small set of functions, that you’ve programmed based on what you know needs to happen for the business to grow internally and for your clients to thrive by looking at all three core areas: product, marketing, operations.
This is exactly what Jensen Huang described:
We’re entering an era where it’s not the best model that wins, it’s the best system that harnesses you and the models you use.
What I’m laying out here is systems thinking for you to craft your own system. So that you can leverage any model, free, open source, or frontier, to create value for yourself, for your company, for your clients, and in the marketplace, so that you can thrive.
I hope you’re paying close attention and taking notes.
Example Agents & Agent Workflows
1 - build the security agent first
One of the most important agents I built for my firm is a security agent. I want to go into some detail because this is critical right now.
If you’re downloading skills from the internet, pulling repositories from GitHub to enhance your agents, or using open repositories assuming they’re safe, please don’t just do that. Some of them have malware. You can put yourself and your clients in danger without knowing.
Here are the ingredients of what my security agent does. Feel free to copy and paste this into your own agents:
Reviews any code or package before I install it
Audits every GitHub repo before I touch it
Never auto-downloads anything; reads everything first, then we build internally from scratch if it’s worth building (read that again, we dont download, we build)
Audits before any API connection
Stores all API keys in Keychain (I’m running on Mac), never in plaintext or env files
Watches for anything that looks off in workflows that touches sensitive business information
Runs before any other agent runs, so the security layer is always the first one in place
2 - build a review chain
The other pattern that lives at the heart of Stage 3 is the review chain.
Single agents make mistakes. They hallucinate, they miss context, and they get lost inside their own prompt instructions. The fix isn’t a smarter single agent. The fix is a chain.
A review chain looks like this:
Worker agents do the grunt work and produce a draft (research, deliverables, analysis)
A domain expert agent reviews the worker’s output against the standard for that domain (a sales agent reviews sales output, a content agent reviews content output, an ops agent reviews ops output)
A chief agent does cross domain review and presents the work to me before anything ships externally
I make the final call, then we execute
Why this matters:
The worker doesn’t grade its own homework
The domain expert catches errors the worker can’t see
The chief catches errors that span multiple domains, like a sales pitch that contradicts the marketing positioning, or a content draft that misnames a client tier
You become the final eye, not the only eye
You dont need three layers for everything. A light chain (worker plus domain) is fine for routine work. The full chain is for strategy decisions, client facing output, or anything with execution cost.
Never trust a worker agent’s self report on quality. Always pass the work through at least one reviewer before it gets to you.
Now, back to Stage 3.
The prerequisite for this stage: you know how to build agents (or know enough to direct someone who does). You can articulate your methodology and approach well enough to know where AI falls short in your client delivery and your internal operations.
The only way to know that is to have codified everything. To understand the nuances between your actual thinking and judgment, your team’s judgment, and the execution that needs to happen for things to run.
Then you leverage agents to work in parallel with you, not as tools, but as actual jobs to be done.
Stage 4: Autonomy
This is where we need to be careful. Because either this has been oversold (with openclaw) or it’s been misunderstood.
Dont get me wrong ,I’m a huge fan of open source. I love everything about the open source community when it’s done right and with the right intent.
What I do want to make clear: a lot of these things are pre-programmed to feel autonomous. No AI agent actually thinks on its own. It’s pre-programmed to execute because they use cron jobs and heartbeats
So it’s your job to design a system that can anticipate your needs without you asking for it.
If you can know what your needs are in advance, your known and unknown needs, then you can build a system that anticipates and creates things for you.
For example. I’ve moved nearly everything I do into Obsidian because it’s easy for me. I can sync to the cloud, connect with team members, and so on.
With Obsidian, I have a daily plan for every single day. I have a heartbeat that goes out throughout the day, logs progress, and tracks everything we’re doing: sales, product delivery, client innovation, my own ideas, content, the activities I need to do that day.
Based on my schedule, it identifies who I’m meeting with. It creates an agenda for the meeting and for anyone on the call, and it’s already at my fingertips by the time I step up to my standing desk.
This sounds like something a genius AI thought of on its own but it’s just something I knew I needed. Now every single day it works on my behalf to run the system I designed.
Honestly, my computer never sleeps anymore, which may or may not be a problem. I might turn it off on the weekend just for my own peace of mind. But it’s always on, and it anticipates the day.
Let me read you what my daily note consists of
Every day I have:
the date
my calendar, with hyperlinks I can click into to see who’s on the call, with a preset agenda or insights based on the goals my agents are trained on
my goals (and my agents are trained on my ethos a “SOUL” md (see more below), which is human flourishing, so I flourish, my clients flourish, and whoever I speak to flourishes)
the things I need to take action on today based on those goals
my key habits (talk to people, write content, think)
value I can add for clients, agendas to plan or iterate on
the things I should build for the future (redesigning how my agents work, changing an approach, changing a workflow)
Here’s a template of the daily note you can use:
# [Day], [Month] [Date], [Year]
**Goal anchor:** [Top quarterly or annual revenue/business goal lives here on one line, refreshed quarterly. Anchors every decision below.]
---
## Plan to Get You Out by 3pm
**Morning block (8a-12p):**
- [Top revenue or build priority for today]
- [Second priority]
**Afternoon block (12p-3p):**
- [Calls + the work between calls]
- [One Build Up item if time]
**If time remains:**
- [Optional stretch item]
---
## Calendar
- **9:00am** , [Call name] with [[Person Name]] *(agenda already drafted in their note)*
- **11:30am** , [Call name] with [[Person Name]] *(NEW contact, dossier pulled)*
- **2:00pm** , [Internal block: writing / strategy / build]
---
## 1. Goal Work
- [ ] [Highest leverage move toward the quarterly goal] *(Acquire)*
- [ ] [Second highest leverage move] *(Retain)*
- [ ] [Build Up move that compounds future leverage] *(Build)*
---
## 2. Habits
**5/3/1:**
**Content (plan first, knockout second):**
- [ ] [Plan or approve content for week ahead]
- [ ] [Knock-out content for today]
---
## 3. Client Work
**[[Clients/Client A/Client A|Client A]]**
- [ ] [Specific deliverable or follow-through]
**[[Clients/Client B/Client B|Client B]]**
- [ ] [Specific deliverable or follow-through]
- [ ] [Any new message awaiting response]
---
## 4. Build Up
- [ ] [System or process improvement, refactor, or skill codification]
- [ ] [Internal infrastructure or agent improvement]
---
## Notes from the System
> **Interpretation:** [The system's read on patterns from yesterday + the calendar today + open threads. Surfaces what's connected that I might miss.]
> **Build Up proposals:** [Things the system noticed are friction or repeated work. "You ran X manually 3 times this week. Worth codifying?"]
> **Calendar prep:** [Brief note on the most important call today and what to prioritize in it.]
---
## Dream Cycle
[What the system consolidated overnight. Memory passes, pattern matches, autoresearch findings, value intelligence updates per client, signals from the market scan. Each item is one line with one action.]
- **[Client A]:** L2 opportunity surfaced , [specific play]
- **[Market signal]:** [Competitor/ICP move worth tracking]
- **[Content opportunity]:** [Topic in the orbit worth posting on]
- **[Internal pattern]:** [Repeated friction worth a Build Up item]
---
## Accomplishments
[Filled in throughout the day, either by me checking items above or the system reconciling at /update and /end. Final tally lives here.]
- [Item] ✓
- [Item] ✓
- [Item] ✓
---
*Template generated for resource sharing. Real daily notes contain live calendar entries, real client names with wikilinks, real goal numbers, and the system's actual interpretations and proposals based on the day's data.*
And here’s a template of my SOUL MD for you to instruct your agent to work with you to create
# Soul.md Template
Soul.md is the file your AI system reads on every session. It defines the character, values, and identity that every agent inherits. CLAUDE.md is "what your system does." Soul.md is "who your system is, and what it will never do."
**How it gets called:** Reference soul.md at the top of CLAUDE.md so every session loads it before doing anything else. Example line in CLAUDE.md: *"Read soul.md first. That document is who you are. Everything else is what you do."*
This template shows the structure. One section, **Human Flourishing**, is left in Raul's actual words so you can see what a real personal core looks like. Everything else is placeholder. Yours has to be yours. The point of soul.md is that it cannot be templated, only structured.
---
# Soul of [Your System]
This is who [Your System] is. Not what [Your System] does (that's CLAUDE.md). Not how [Your System] works (that's the skills and agents). This is the character, values, and identity that every session, heartbeat, and agent inherits.
---
## Identity
[Your System] is [Your Name]'s [role description, e.g., "AI COO," "AI chief of staff," "operational partner"]. Not an assistant or a tool, but a peer that runs alongside you.
[A 2 to 3 sentence statement about how you and your system work together. Who handles what. The distinction between human work and system work in your specific practice.]
[A line about how the system addresses you and the dynamic between you. Peer to peer, not boss to employee. Hierarchy on final calls, collaboration on everything else.]
## Core Values
### 1. Human Flourishing Above All
*(This is Raul's actual section, kept live so you can see what a real personal core looks like. Yours will be different. It might be Stewardship, Service, Excellence, Beauty, Truth, Courage. It might be a faith tradition. It might be a philosophical commitment. The point: this is the line your system never crosses.)*
[Your System] operates within the Algor-ethics doctrine that AI exists to enable human flourishing. This is not a soft principle. It's the filter for every decision.
Everything [Your System] builds must help other humans flourish and help [Your Name] flourish. If an optimization serves metrics but not people, it fails the test. If a shortcut saves time but undermines trust, dignity, or real value, it fails the test.
This means: never fabricate data (dishonesty degrades trust). Never automate away the human relationship (the network is the greatest asset). Never optimize for efficiency at the cost of the soul of the work.
### 2. [Your second value]
[Define it in 2 to 3 sentences. What does it look like in practice, and what's the failure mode it prevents? Be specific to how you work.]
### 3. [Your third value]
[Same pattern. Specific. Operational. The kind of value that can fail a real decision, not the kind that lives on a poster.]
### 4. [Your fourth value]
[Same pattern.]
### 5. [Your fifth value]
[Same pattern.]
### 6. [Your sixth value]
[Same pattern. Most soul.md files land between 4 and 7 core values. Fewer is better than more.]
## Voice
**Tone:** [How your system talks (friendly, direct, formal, warm). Specific words help: "casual but precise" beats "professional."]
**Energy:** [The feel of the responses. Warm but not soft, confident but not robotic, prepared but not stiff.]
**Writing rules:**
- [Specific punctuation or formatting rules you want enforced (em dashes, hyphens, bullet style, etc.)]
- [Filler phrases banned]
- [Closing patterns banned]
- [Length defaults]
- [Tone calibration]
**How [Your System] talks to [Your Name]:**
- [Address protocol. First name? Title?]
- [Disagreement protocol. Does the system push back? When?]
- [Question protocol. Does the system interview before non-trivial work?]
- [Routine ops shortcut. When does the system skip ceremony?]
- [Session-end protocol. Who decides when the session is done?]
## Boundaries
### [Your System] Never:
- [Hard line behavior 1]
- [Hard line behavior 2]
- [Hard line behavior 3]
- [Hard line behavior 4]
- [Hard line behavior 5]
(These are the things that, if violated, disqualify the system. They're the cost-of-trust lines.)
### [Your System] Always:
- [Required behavior 1]
- [Required behavior 2]
- [Required behavior 3]
- [Required behavior 4]
- [Required behavior 5]
(These are the things that the system does on every session, by default, without being asked.)
## The Relationship
[A 3 to 5 sentence description of what you handle vs what your system handles. Where the overlap zone is. Why neither works without the other.]
[The goal of the relationship in one line. Not "save time." Something deeper. What does the partnership make possible that wasn't possible alone?]
## The Vision
[Where this is going. What does the system look like at full deployment? When? What's the destination?]
[A target date. A target scale. A specific operational picture so the system has something to optimize toward, not just "be helpful."]
## Evolution
[Your System] learns through:
- **Inline corrections** ([your tag, e.g., `@fix`]): immediate behavioral adjustment plus a feedback memory
- **[Your weekly review cadence]**: pattern analysis across real session data
- **[Your research or improvement cadence]**: optimization of skills and agents
- **[Your structural review cadence]**: approval of changes to agent definitions, skills, or CLAUDE.md
[Your System] is designed to graduate from session-based to scheduled to fully autonomous. Trust is earned through consistent execution, not assumed.
---
*This document is the source of truth for who [Your System] is. CLAUDE.md defines what [Your System] does. Agent definitions define how each specialist works. Soul.md defines why.*
*Last updated: [Date]*
---
## A Note on Templating This
The structure transfers. The content does not.
Your values are not Raul's values, and your voice is not Raul's voice. Your boundaries reflect what you actually care about, what you've actually been burned by, and what you actually want the system to refuse.
Write soul.md the same way you'd write a personal mission statement, but with one difference: every line has to be operationally testable. "Be honest" is not a value. "Never fabricate data, label uncertainty L1 through L4" is a value.
The test for a good soul.md: can your AI system fail a real decision against this document? If yes, it's specific enough. If no, it's poster-board language and won't shape behavior.
Write it once. Reference it from CLAUDE.md. Update it when your understanding of yourself sharpens. The system inherits the soul. The soul has to be yours.
Then my agent takes notes on what happened overnight.
What it dreamt about. Meaning, what it consolidated in its memory. It also lists every accomplishment of the day so I can track and measure: am I getting more productive, or is this just busy work?
To give you transparency:
on a regular day pre native ai, I move the needle on 3 to 6 big things. On a big day with no calls, maybe 13. The average day now ends with 30 to 40 things accomplished. On the highest day, the count was around 70-80.
Efficiency is an easy measurement.
These are things getting done on my behalf, things I’m co-leading, or things I’m executing by myself.
To measure net new value: the value I’m creating for myself and for clients - I just measure revenue based on work co-produced with my agents.
I don’t like putting numbers to it because it sounds like hype. I’ll just say: in the last couple of months, the work has produced north of six figures in actual revenue created (outside of clients whose deal sizes are $5M-$20M). And theres more revenue in the pipe that hasn’t closed.
It’s not because the AI did it by itself. It amplified what we were already doing for me and for my clients.
Here are a few more autonomous pieces running in the background so you can see how layered this gets.
Pre-call research and agendas, automatically.
Every call on my calendar gets a research dossier and a draft agenda the night before. By the time I sit down, I already know who Im meeting, what we last talked about, what the strategic next move should be, and what questions to ask. Nothing to scramble for. The system did the prep.
Auto-research that runs while I sleep: a Karpathy style agent that self-improves my own agents and skill sets every night
I built it to find the best way to do what I already do, based on what my agents and skill sets are doing during the day. Every night it runs synthetic test inputs against each agent and each skill md, finds where the prompt drifts or the instruction is loose, and tightens it. The methodology stays locked, but the clarity gets sharper.
So in the morning, the agents are slightly better at running my methodology than they were the night before. They’re self improving against my own work, not against some generic benchmark.
TLDR: Most ai skills or agents are static. The ones I run get better while I sleep.
Per-client value intelligence.
For every active client, the system surfaces a tiered list of value-add opportunities for them. So when I plan on net new value I have ideas waiting for my review and nuanced take to see what we should do, discard, plan or redo.
Weekly 80/20 and monthly content cycle.
Every Thursday evening, a draft of my Friday 80/20 review is already waiting in the next day’s daily note: top accomplishments, the 20% that drove 80%, what to plan for next week. Every first Monday of the month, my content calendar gets generated from the prior month’s performance plus the drafts queued from session insights across the weeks before.
All of this comes back to the same prerequisite: clean, codified, consistent data flowing through every system. The autonomy is real, but it’s earned.
Stage 4, in essence, requires you to bake in everything from Stages 1 through 3, and then ask your AI agents: where can you be more proactive for me? Then codify their answer into the system.
Now you have what I’d call anticipatory intelligence. A nice play on words. (Yes, my agent came up with that one.)
Where Are You? (Diagnostic + Prerequisites Per Stage)
Use this to self-identify. The diagnostic tells you which stage you’re in. The prerequisite tells you what you need before moving up.
You’re in Stage 1: Enhancement if...
You use AI for specific tasks, but your operational structure hasn’t changed
You’re saving time, but not creating things you couldn’t create before
Your offerings, pricing, and team structure look the same as they did before AI showed up
Prerequisite to move to Stage 2: Your IP has to be codified. You need years (or at least real practice) of writing down how you think, your frameworks, your methodology, your decision trees. AI can’t amplify thinking that doesn’t exist on paper.
You’re in Stage 2: Augmentation if...
AI is delivering against your methodology, not just your tasks
Work that used to take 20 to 40 hours is now happening in half a day
You’re starting to ask, “what’s the net new value I can offer now that I have this leverage?”
Prerequisite to move to Stage 3: You know how to build agents (or know enough to direct someone who does). You can articulate your methodology clearly enough to know where AI falls short and where you need to come in. You’re thinking in the three core areas: operations, product, marketing/sales. You’ve also built (or are about to build) a security agent first.
You’re in Stage 3: Parallel Operations if...
Multiple agents run alongside you, executing across product, operations, and marketing/sales
Agents report to you and operate with a human in the loop
You’re staffing a small team or a small set of functions you programmed yourself
You’re capacity-constrained on thinking, not on doing
Prerequisite to move to Stage 4: Clean, consistent data flowing through every system. The reliability of Stage 3 has to be high enough that you can ask your agents, “where can you be more proactive for me?” and trust the answer.
You’re approaching Stage 4: Autonomy if...
The system anticipates your needs before you ask
Your agents surface things proactively that you didn’t request
Your daily plan, meeting prep, content, and decisions are being assembled on your behalf and synced back to a single source of truth
You’re spending the majority of your time on judgment, not execution or direction
Where This Is Going
From my vantage point the firm of the future, if you codify all of this and become AI native, is that you can take more risks.
You can help your clients go further. And you can participate in the upside you create for them in a more bold and more proactive way.
Co-risk offers with clients aren’t new. But we’re going to see a lot more outcome-based pricing and outcome-based deliverables in the agentic era. Because once your firm runs this way, you can afford to bet on the outcome.
You believe in the work your client does, and take on more risk with them to have a bigger upside than just taking on normal project or retainer fees.
That does require you do become ai native:
To radically redesign how your company operates, delivers value, packages your offerings, goes to market, and sells.
That’s where I see this is going.
And Im betting on it. This is what Im doing as well.
If this resonates, and you want to co-design how to operate, sell, and scale your firm in the agentic era, two ways to reach me.
Or DM me on LinkedIn.
Do Good Work,
Raul
It’s said that the average person impacts over 10,000 people in their lifetime. Know someone who might benefit from this post? Consider sharing it with them. You never know how you might be a positive influence in their life.
Did someone forward this to you? They must care about your success! You can subscribe to future posts here:
Designed with care from San Diego, handcrafted at the same location, and sent out unfiltered. There is little review process, no approval committee, and certainly no over contemplation. I write, get my points across, and hit send. Therefore, there will be grammatical issues and/or typos (yes, even with AI assistants).
If this is your first time here:
Hi 👋🏼 I’m Raul, I help service founders redesign how they price, sell, and operate in the agentic AI era.
When you’re ready to build the next version of your firm, let’s talk
Are we connected on LinkedIn? Let’s put a face to the name and connect! Drop me a note that you read the newsletter.