You already made
the investment in AI.
Let’s go get the return.
I'm Cindy Au. I bring harmony to the chaos of AI implementation for small (100-3000 person) pharmaceutical or other highly regulated manufacturing sites who have already taken the first leap into the streams of the future but may not be seeing productivity gains.
I design the workflows between your people and your technology so the money you have already spent on digital tools and agentic AI starts producing.
Most AI implementation consultants hand you a plan and leave. I start with the conditions you have, the people you have, the tools they use, and the pressures affecting your people, and I design the flow from there.
Planning the ideas is the easy part. Building a sustainable implementation is what will get you ahead of your competitors. I help you get it done.
This is the future of work. Are you ready?
Just a conversation about what you have built and where it has stalled.
AI implementation consulting helps a company move from buying AI tools to actually using them. You have the business strategy. My job is finding the root causes of a stalled implementation: your operating conditions, your workflows, and the people who have to run the systems. That is what decides whether the money you spent produces a return.
Your people never got the time to learn what you bought them.
The tools landed. The deadlines did not move. So your team is still expected to hit quarterly production goals while using systems that change the way work gets done. But they haven't been given any more time to learn the whys behind the how. They're just pushing buttons, and if something goes wrong, they haven't had time to learn how to troubleshoot or what might have caused the issue.
And while the reasons for cutting out time for learning seem reasonable, "They are adults. It is their job. They are paid to learn," a piece of the puzzle is missing. Certainly they are adults and they will learn. Just not on the same timeline, with the same headcount, for the same output you may think is possible.
So the work routes around what nobody understands. Information stops moving between groups. The handoffs become the place where everything stalls.
Why your AI implementation hasn’t delivered ROI yet
The bill for this arrives later, and it arrives bigger.
So you've brought AI on board, sunk in the time and money to pay for the technology, but the outcomes are negligible. Where is the value? The cost you can see now is the spend on tokens, tools, and agentic AI subscriptions that seem to be yielding results. Project plans laid out in days, not weeks. Budgets drawn up in hours, not days. Memos and communications in seconds, not minutes.
Yet the cost in tokens for LLMs doing the work of humans is actually more than the cost saved in human salaries.
The roles AI absorbed first were the entry-level roles, and those were where people learned how the business actually fits together. You will not feel that this year. You will feel it in three years, when you need a senior manager and realize the jobs that used to produce one no longer exist.
Then there is the cost for plans that simply do not get sustained, because the workflow never changed to accommodate the shift in job structures with agentic AI and digital tools. Six months later everyone is back to the old workflows with a more expensive tool sitting on top of it, and no one in management who has time to learn the ways digital and agentic AI have altered the responsibilities of humans.
Your people were the part missed in the design process
I am not here to talk you out of AI. I use it every day and I think you were right to build on it. What I have watched go wrong is almost never the technology. It is the fact that nobody designed the space between the technology and the people who have to run it.
That gap is where I work. Most AI transformation consulting stops at the strategy layer. I stay through the implementation instead of handing you a plan at the door, which is why this is AI enablement consulting rather than AI strategy. Before anything gets rebuilt, I look at the conditions you are operating in: who your people are, what is already on their plate, where the work moves between groups, and why you have come to this point of needing something to change.
I have sat in the workflow version that does not work. A big firm arrives, delivers something polished, hands it over, and says learn it. Six months later nothing is left. What survives is what gets designed around the people who were always going to be running it.
Start with what is actually happening.
AI consulting services can cover a lot of different elements depending on who is selling them. Here are the three stages of my AI implementation services that allow a working relationship with clear entry and exit points.
ASSESS THE CONDITIONS
Before I design anything, I need to see what is really going on. Not the org chart and not the rollout plan, but where the work stalls, who is carrying what, and how much room your people genuinely have. You come out of this with a clear map of your gaps.
DESIGN THE FLOW
I create a prototype of how the work could move with your people and your technology in it, and with the time to learn actually accounted for. You see exactly what changes, who it touches, and what it will take before anything gets built.
BUILD AND PROVE IT
We implement it in stages over ninety days with a small cross-functional team, then measure what changed. Not a pilot that quietly gets shelved. A working version, running in your real conditions, with a report at the end telling you what it produced.
It starts with a thirty-minute conversation, at no cost. Tell me what you have already built, what you expected it to do, and where it stopped. I will reflect back to you what I am hearing for your pain point, and you'll tell me if it matches your understanding, because those two are often different.
If you believe I've accurately summarized your key issue, the work rolls out in three stages: an assessment that tells you what is actually happening, a prototype design so you can see connections and accountabilities before you commit, then the build, which puts it into real conditions and measures what changed. Let's dig deeper.
The process:
ASSESSMENT, DESIGN, and IMPLEMENTATION
A working session with you, and a written read on the conditions you are actually operating in. I look at where the work moves between groups and where it stalls, what is already on your people's plates, what has been tried before and did not hold, and why this has become urgent now. You come out knowing where your gaps are and what it would take to close them. The read is yours whether or not we go further, and if you go on to a build, the cost is credited toward it.
OPERATING ASSESSMENT
01
Before anything gets rolled out, I design a prototype of how the work could move, with your people and your technology in it, and with the time to learn built into the schedule instead of assumed. You see the redesigned workflow running in practice, who it touches, and what it will take, before you commit to building it.
90 DAY STAGED IMPLEMENTATION
02
We build and test it over ninety days in three stages with a small cross-functional team, working in your real conditions rather than in a demo. At the end you get a written report: what changed, what it produced, and what I would do next.
PROTOTYPE DESIGN
03
I work with cross-functional teams of six to ten. More than that and there are too many voices with whom to cooperate, coordinate, and collaborate.
In pharmaceutical operations and other regulated industries, none of this happens outside your compliance obligations. Every workflow change has to hold up against validation, change control, and audit requirements, so I adjust the redesign to fit those constraints rather than around them. When a system changes how work gets done, the documentation and the training records have to change with it, and I build those factors into the schedule instead of leaving them for later.
This works when you are willing to look at more than one version of the answer.
I work with leaders at mid-size companies, most often in pharmaceuticals and other regulated industries, who have already put real money into AI yet are not seeing the expected outcomes. The pieces all seem to be present: equipment, digital and AI-enabled tools, and the people. What you do not have is a design that connects them all.
Most companies your size fall into a gap. Too big for one freelancer to carry, and too small for the large firms to pay real attention to. That gap is where I built this.
This is also for you if you have not started yet and want to avoid the costs of failures that a large company can absorb but a small to mid-size company cannot.
The one thing I need is a leader who is open to hearing more than one perspective on their own organization, including the parts that are uncomfortable. If a plan is already decided and what you want is someone to deliver it, I am not the right person. If you want to know what is actually happening on your floor before you commit the next round of budget, that is exactly what I do.
Cindy Yee Au
Twenty-five years of watching how people actually learn, and three of them on a production floor.
I spent three years as an instructional designer at a rare disease pharmaceutical manufacturer, and because of how that site was built I saw the production line end to end, which most people in my role never do. That is why I can usually tell you where your rollout will break before it breaks. These failures are not random. They cluster in the same few places every time: the handoffs between two groups, the team nobody consulted, and the step somebody assumed would just get absorbed.
When a company was struggling with qualification time for employees to get on the floor for aseptic operations, I helped redesign the process and brought it from an average of six months down to two. When a new drug substance tech transfer hit the floor, I found the bottlenecks in the path preparing people to run a 17-step manufacturing process by uncovering and providing the necessary 32-step technical prep that 100 people working across five functions needed to produce results, providing a 60% increase in visibility for training.
Before that came two decades in classrooms and school administration, and an M.Ed, designing how people learn with no time, no budget, and a deadline that does not move. It is why I can put a number on what your implementation plan is currently guessing at: how long it actually takes your people to get competent, and what has to come off their plate for that to happen. Executing a redesigned workflow takes a guided team. I am that guide.
Questions I get asked.
-
Then you probably have a good plan. The question is whether it survived past the handoff. What I do differently is design around the people who have to run it, and stay through implementation, which is where most plans quietly stop getting followed. If your plan is solid and simply is not sticking, that gap is exactly where I work.
-
Because time is the constraint, not content. Your people are not short on material. They are short on room. Training that gets added on top of unchanged production targets does not get absorbed, and you end up paying for it twice.
-
You are not spending more on the technology. You are spending on the thing that makes the technology produce. The money you have already committed is either going to return something or it is not, and right now the deciding factor is not the tools.
-
The first conversation is thirty minutes. The assessment happens inside a couple of weeks. A full build runs ninety days in three stages, and you see what each stage produced before the next one starts.
-
They usually do not, once they realize someone is asking about their conditions rather than adding to their workload. Resistance is almost always a signal that people have been asked to absorb something impossible. That is information, and I use it.
-
Pharmaceutical manufacturing is where my most recent experience is, and regulated industries generally are where I understand the pressures fastest. But the work applies anywhere the technology arrived faster than the people could be prepared for it.
-
Usually because the tools got implemented and the conditions around them did not. The technology works. What nobody planned for was the time your people needed to become competent with it, and the handoffs between teams that the new systems quietly changed. This is the enablement side of AI rather than the technology side, and until it is designed, the tools run at a fraction of what you paid for.
-
In my case, three things. Assessing the operating conditions you have right now. Designing a prototype of how the work should move, with your people and your technology in it. Then building and testing it over ninety days in your real environment, and measuring what changed.
-
When you are evaluating an AI implementation partner, three questions are worth asking anyone you consider. Do they start from your actual operating conditions or from a template. Do they stay through implementation or hand you a plan and leave. And can they tell you what they will measure before the work begins. If those answers are not clear, you are buying a document rather than a change.
-
We agree on what we are measuring before the build starts. It is usually how long it takes people to reach competence, your error and deviation rates, and the deadlines you stop missing. At the end of the ninety days you get a written report on what changed and what produced it.
You have already made the harder decision. This is the one that makes it pay.
You committed to building on AI, which took nerve, and you are not wrong about where this is going. What is missing is the design that connects it to the people who have to run it.
Start with a conversation. Tell me what you have built and where it stopped. I will give you an honest read on whether a workflow redesign will help, or whether you need to start in another place before I can make a difference for you.