Change management
New tools fail when people are handed them cold. I work inside your team until the new way of working is the normal way of working.
AI change management and FDE services
AI that works inside your organization, run by your own people.
I come in and map how the work actually happens. Together we identify what to improve. Then I implement it with your team: automating the repetitive parts, keeping your knowledge private and secure, and teaching your people to run it. Then I leave you in charge.
What you get is the work done, and your own people able to keep it running.
- Change management
- Institutional knowledge
- Streamlining
- Automation
- Security sweeps
- Multi-agent coordination
- Education
For organizations
Bring it into your organization
Hands-on implementation, change management, and onboarding for your team and your customers. Healthcare, energy, education, agriculture, finance. Read this door ↓
For individuals and self-starters
Run it yourself, free
The open-source kit I run my own life and work on, eighteen short tutorials, and a community as more people build. Read this door ↓
What I bring
Seven things. The seventh is the point.
Institutional knowledge
Decisions, procedures, and handoffs in one searchable place your whole team can use.
Streamlining
Before anything gets automated, we map how the work really happens: who does what, where it stalls, and what should not exist at all.
Automation
The reports, research, and routine work that eat your week get done by AI assistants, one process at a time, and stay done.
Security sweeps
Scheduled checks that passwords, dependencies, and data stay where they belong. Boring on purpose, and running whether anyone remembers or not.
Multi-agent coordination
Several AI assistants working as a team with defined roles and one shared memory, instead of one chat window doing everything badly.
Choose an offering to explore.
07 · The point
Education
Your people learn to run and extend the system themselves, and I help you teach your customers the new way too. The goal is more agency for you, not a dependency on me.
The system behind the offer
Agent OS, in a nutshell.
This is not a slide. It is the system I run every day, and the one I set up for clients. Every piece below is real and running.
A centralized brain
Every decision, handoff, and procedure is written down in one place (an Obsidian vault) that every AI assistant and every person can search.
A team of assistants with a command center
Assistants running on different AI models coordinate in one group chat. Each has its own lane, its own instructions, and its own project board, so work does not collide.
Routines that run without a model
Nightly backup, security sweeps, credential checks, supply-chain checks, and message relay run on schedule with no AI involved at all, which makes them cheap and predictable.
Spend you can see
Every assistant’s usage is tracked. Failed tasks and wasted tool calls count as wasted money, and an independent grader reviews it weekly.
Maker and judge
One assistant does the work, a second one reviews it. Blind spots get caught before a human ever sees the result, so you are not the only reviewer.
One shared brain. Five connected parts.
Freedom
Nothing here locks you in.
The system is built around your work, not one vendor. Every piece under it can be swapped; we choose the models, knowledge tools, and machines together, and check what needs adapting before making a change.
Built to improve, with checks at every step.
The work, the security checks, and the way we judge quality can all be refined. Agent OS is the product, and it keeps evolving: when a new model ships, it reviews the whole system and proposes what to improve, and that has happened with every release so far. Every proposed change is tested, reviewed separately, and approved before it goes live. If it makes things worse, we keep what worked.
The work gets a second look.
- What prompts a change
- A repeated bottleneck, a failed task, or a useful new model.
- What gets tested
- A revised process, checked against the task it needs to do.
- What can move forward
- A reviewed version, with the change recorded.
The checks can get sharper.
- What prompts a change
- A new risk, a missed warning, or a change in how you work.
- What gets tested
- A check that catches the risk without breaking approved work.
- What can move forward
- A reviewed check that fits how you already work.
The review can learn, too.
- What prompts a change
- A mistake that escaped review, or an unclear standard.
- What gets tested
- A revised test that catches the miss and still recognizes good work.
- What can move forward
- A clearer test, checked by a separate reviewer.
The same path, every time
- 01Propose
a change - 02Test it on
the work - 03Independent
review - 04A person
approves
Doesn’t pass? Revise it, or keep what worked.
A review process, not a promise that every change is an improvement.
The maker does the work. A separate reviewer checks it. A person approves what goes live.
A system that can keep learning
Built to improve, with checks at every step.
The work, the security checks, and the way we judge quality can all be refined. Agent OS is the product, and it keeps evolving: when a new model ships, it reviews the whole system and proposes what to improve, and that has happened with every release so far. Every proposed change is tested, reviewed separately, and approved before it goes live. If it makes things worse, we keep what worked.
The work gets a second look.
- What prompts a change
- A repeated bottleneck, a failed task, or a useful new model.
- What gets tested
- A revised process, checked against the task it needs to do.
- What can move forward
- A reviewed version, with the change recorded.
The checks can get sharper.
- What prompts a change
- A new risk, a missed warning, or a change in how you work.
- What gets tested
- A check that catches the risk without breaking approved work.
- What can move forward
- A reviewed check that fits how you already work.
The review can learn, too.
- What prompts a change
- A mistake that escaped review, or an unclear standard.
- What gets tested
- A revised test that catches the miss and still recognizes good work.
- What can move forward
- A clearer test, checked by a separate reviewer.
The maker does the work. A separate reviewer checks it. A person approves what goes live.
The same path, every time: Propose a change, Test it on the work, Independent review, A person approves. Doesn’t pass? Revise it, or keep what worked.
How we work
Self-improving, reviewed, measured.
Every change to the system goes around the same loop, one small change at a time, so nothing breaks in a way you cannot trace. Nobody grades their own homework.
Choose a step to explore the loop.
Step 1 of 7
Plan the change
Safeguard at this step
One change, written down first.
Each change is small and recorded before it is made, so we know exactly what to check afterwards.
Six safeguards, on every change.
The ones lit up apply to the step you chose.
The maker never signs off.
The assistant that does the work never certifies it.
An independent judge reviews.
A second assistant checks the result before a human ever sees it.
Memory keeps what was learned.
Decisions and failures are written down where the next run can find them.
A gate blocks regressions.
Tests check that a proposed change does not break the work we already rely on.
A security check runs on anything sensitive.
Credentials, data, and dependencies get looked at before anything changes.
A person approves what goes live.
You or me, every time.
- The maker never signs off. The assistant that does the work never certifies it.
- An independent judge reviews. A second assistant checks the result before a human ever sees it.
- Memory keeps what was learned. Decisions and failures are written down where the next run can find them.
- A gate blocks regressions. Tests check that a proposed change does not break the work we already rely on.
- A security check runs on anything sensitive. Credentials, data, and dependencies get looked at before anything changes.
- A person approves what goes live. You or me, every time.
For organizations
Hands-on work, inside your walls.
This is the forward-deployed part (the FDE in the tagline): I sit with your team, automate the reports and research that eat their week, set up the system, and onboard the people who will run it. Most AI efforts stall on people’s time and unfamiliarity, not on the technology. That gap is the job.
One system, scoped around your team’s size, sector, tools, and constraints. Sectors I work in, with case studies added as each one lands:
- Healthcare
- Energy
- Education
- Agriculture
- Financial services and fintech
Map
A short conversation.
What eats your team’s time, what data you have, who would run this.
Improve
Identify ways to improve.
Once we map it, we identify how we can improve this, and that is what we implement.
Implement
A scoped first project.
One process, automated end to end, with your people trained on it.
Teach
Onboarding and education, inside and out
- Employee onboarding and new rollouts, inside the tools your people already use (Microsoft Teams, for one).
- Customer onboarding, for when the way you serve customers changes because the way you work did.
- Just-in-time guides, built from your own material by AI assistants: short tutorials, plus quick onboarding checks that show each person where their gaps are, so nobody sits through what they already know.
Map. Improve. Implement. Teach.
Your people have the deep understanding of the company, the customers, and the processes. I bring the skill set and the system, working with the leaders of the team (cross-functional when it needs to be), to cut the learning curve and the implementation curve.
Inside your own AI environment, if you prefer.
The system is a set of files, instructions, and procedures. It can run inside your own AI environment, set up with your team on your terms.
- Map: A short conversation. What eats your team’s time, what data you have, who would run this.
- Improve: Identify ways to improve. Once we map it, we identify how we can improve this, and that is what we implement.
- Implement: A scoped first project. One process, automated end to end, with your people trained on it.
- Teach: Onboarding and education, inside and out
- Employee onboarding and new rollouts, inside the tools your people already use (Microsoft Teams, for one).
- Customer onboarding, for when the way you serve customers changes because the way you work did.
- Just-in-time guides, built from your own material by AI assistants: short tutorials, plus quick onboarding checks that show each person where their gaps are, so nobody sits through what they already know.
Your people have the deep understanding of the company, the customers, and the processes. I bring the skill set and the system, working with the leaders of the team (cross-functional when it needs to be), to cut the learning curve and the implementation curve.
Inside your own AI environment, if you prefer. The system is a set of files, instructions, and procedures. It can run inside your own AI environment, set up with your team on your terms.
For individuals and self-starters
The same system, free, for your own life and work.
I run my own life on this: my writing, my publishing, my research, my analytics, my creative projects. The starter kit is open source and free, the tutorials walk you through the first setup in eighteen short videos, and a community is coming as more people build.
Questions people ask
The practical part.
- How does pricing work?
- Per project, agreed before the work starts. A small first project lets you judge the value before committing to more.
- How will you know it is useful?
- Before I build anything, we agree what a useful result looks like and how we will check it, against your organization’s goals rather than for the sake of measuring. Two sides get counted. What we did: the process before and after, and the tokens and model spend on our side. What that did for you: the time given back, and what it went to, tied to a goal you already track.
- How long does it take?
- How long depends on the process, and we agree on it before we start.
- Where does the work happen?
- Inside your organization, alongside your team, in the tools they already use. Where it runs is your choice; see Freedom above.
- What about our data?
- It stays yours, inside a setup we agree on together; see Freedom above.
- What happens after you leave?
- Your people run it, with everything written down so they can change it. If you want help improving it later, we scope that together.
Who you are working with
Yacob Berhane.
I have built and run companies before this. At Pariti, the company I founded and led, we grew a talent community of 70,000+ people across emerging markets. I run Agent OS every day on my own writing, research, and operations, and I set it up for clients the same way. CNBC Africa and TechCrunch have covered Pariti.
70,000+
people in the community we built at Pariti