Varick leads AI transformations helping companies unlock nine-figure efficiency gains by redesigning how work gets done.
by
Varick leads AI transformations helping companies unlock nine-figure efficiency gains by redesigning how work gets done. By the end of this article, you should understand how a business should be rebuilt from the ground up around AI to unlock value at this scale.
You'll know how to identify which workflows in your business are worth automating, and how to redesign them in a way that doesn't disrupt operations.
Follow this account for more break downs and case studies of how AI is being adopted at the largest companies across America. If this work excites you, we're hiring across the board - join us at varickagents.com/careers.
Intro to Transformations
You will not transform your company without rebuilding operations from the ground up.
The industrial revolution taught us that productivity gains don't show up if you don't do this. For 30 years, factories swapped out the steam engine with electric motors and saw little financial benefit. Old factories were built around one central steam engine in the basement that powered every machine in the building.
When electricity came along, factories just replaced that engine with a motor and changed nothing else. They kept the same building, the same layout, and continued working the same way.
What moved the needle was a complete operational redesign from the ground up around electricity. The real unlock was that electric motors could be small and cheap, so every machine was able to have its own. That meant factories no longer had to be built around a single power source, and they could spread out and put machines in the order that work actually flowed. This led to the assembly line, which created major productivity gains.
Henry Ford figured out that you need to rebuild around the technology to create value in the early 1900s, and we're using that same operational redesign playbook today.
Buying software doesn't get you there
Which brings us back to today. Most companies trying to transform with AI are hoping to buy their way there by swapping their SaaS stack for AI tools. Agentic software seats, Copilot licenses, and no-code workflow builders rarely move the needle on their own, because transformation is not a piece of software you can purchase. It's a structural change in how the business operates, and it starts with the people and the processes that run it.
If the AI does not understand the underlying process, it will not create meaningful value. And if the people who own that process are not brought along, adoption will be weak even if the technology works.
That's why you need to spend a few weeks with teams across the business — from accounts payable, to procurement, to operations — and understand how their work actually gets done from end to end.
You should map every workflow, figure out what the ROI of an agent would be in each particular workflow and how to approach it from an engineering perspective, then choose where to deploy the agents where they'd be a good fit (which we'll go over later).
From there, capture company context (tribal knowledge) and convert it into rules, instructions, and decision logic that the agents can follow.
Doing this with each team is the only way to get the context required to redesign the business around AI, and the buy-in required for the transformation to actually stick.
The Operational Redesign
Once every process is mapped end to end, the next step is deciding which workflows should actually be redesigned around AI. This is the operational redesign.
Take it from an agent company — please do not put agents in every workflow. There is a point where agents create more problems than they solve.
In this section, we'll cover how to create value from an AI transformation and how to not disrupt the business while doing so.
How to create value
A transformation is about redesigning each workflow so deterministic work is automated, judgment work is handled by AI where appropriate, and high-risk, high-judgment decisions remain with humans.
Done correctly, this does more than cut cost. The agents should work to give people better context, and better context helps people make better decisions, faster. Consistently better decisions unlock revenue growth. This means that a proper transformation should yield both topline growth and efficiency gains.
We saw this clearly in a sales transformation for a multibillion-dollar revenue enterprise software company. Too much of the sales process was trapped in busy work, and large deals touched six teams across eleven handoff points. So we ran through the entire process outlined in this article, found the right workflows to automate (either with agents or with scripts), and delivered $25m in value in the first year through margin expansion (revenue growth + savings). Proper transformation isn't just a cost-cutting initiative.
How to pick the right workflow
One of the most important parts of an AI transformation is choosing the right workflow to redesign first. Not every process is worth automating, and not every process is a good fit for agents.
The best workflows usually have a few things in common: high volume, lots of manual effort, fragmented systems, repeated handoffs, tribal knowledge, and clear financial impact.
You're looking for places where work is already happening over and over again, but the process is messy enough that traditional automation hasn't solved it. Think data moving through email, Slack, spreadsheets, portals, and ERP systems.
A good workflow to redesign usually has four traits:
It happens often enough to matter. The process should run hundreds or thousands of times a month, or touch enough revenue or cost that improving it creates real value.
It has repeatable decisions. The work does not need to be identical every time, but it should follow patterns. Agents are most useful when they can learn from past decisions, apply business rules, and route exceptions.
It depends on context spread across systems. The more humans are searching between tools to gather information, the more valuable an agent can be. AI is especially useful when the work requires pulling context from contracts, emails, CRM records, ERPs, documents, and internal rules.
It has measurable pain. You should be able to measure the current cost of the workflow (cycle time, error rate, manual hours, delayed revenue, duplicate payments, approval delays, etc.) before and after deployment.
The goal is to separate the work into three buckets: what can be handled with deterministic automation, what should be handled by agents, and what has to stay with humans.
Agents need to self-improve
Make sure to build human-in-the-loop feedback into the system from the start. During training and shadow mode, humans can approve, reject, or correct an agent's actions. Always log the agent's output, the human's response, and the surrounding context so the system can improve over time.
This makes agents meaningfully more accurate after deployment. In the same sales case, accuracy increased by 10% within a few weeks, and the result was an increase in the amount of work that agents can handle autonomously, which led to higher net dollar value created.
How to not disrupt the business when transforming
You need to ensure that the transformation doesn't create complications and is as cost-effective as possible. A couple of steps we live by:
Don't force massive migrations. Most companies have already spent years moving onto systems like Salesforce and NetSuite. Forcing companies to rip out and replace their software just to adopt AI slows the transformation down and forces teams to relearn the software they depend on. At Varick, we strongly encourage building on top of the systems already in place, whether through APIs or computer-use agents. This avoids expensive data migrations, keeps the workflows the business already runs on, and allows the operational redesign to survive even if the underlying software changes later.
Know the data and keep it segmented. In most workflows, the data that powers the transformation falls into four categories: the system of record, the business rules, the raw intake data, and the feedback or memory the agent accumulates over time. Keeping those layers separate is very important. It means an operations person can update a rule without calling an engineer, and it makes the system easier to maintain and scale after deployment. Your goal is to design a transformation so it can keep running smoothly with minimal intervention after deployment.
Over time, the organization starts to reshape around AI
In the first weeks after deployment, agents run in sandboxes, then in shadow mode alongside humans, and only later in supervised production use cases. As confidence grows, workflows are not just automated but often redesigned and improved.
The goal is to redesign operations where it helps the business move faster and create more value.
When we start with a new company, the first workflows we scope are typically:
Accounts Payable — Invoice automation, GL coding, purchase order matching, etc.
Procurement — Vendor onboarding, supplier scorecards, contract compliance, etc.
Sales — Deal desk routing, CRM enrichment, forecast intelligence, commission calculation, etc.
Operations — Exception detection and routing, allocation optimization, returns disposition, etc.
The company gets used to handing off busy work to AI and owning real work, and efficiency gains start showing up on the P&L in weeks.
This is what the outcome of your transformation should look like if done correctly. The process takes its time because it's the foundation to becoming an AI-native company, but when that happens, financial uplift shows up fairly quickly. Make sure you're building your way there workflow by workflow, function by function. It'll take its time, but good transformations are well worth the investment.
Every lesson in this guide has been learned over dozens of large-scale transformations we've done for clients. If your business is doing over a billion dollars in revenue and wants to see the results our clients are seeing, we're now accepting engagements for July. Find us at varickagents.com
