Over the last few months we have watched consultants use these tools on live projects, at real customers, with real deadlines. So this time we will not predict anything. We will just show you what happened.
Two things have changed since we hit publish.
First, the market caught up to the argument. McKinsey’s 2026 research on ERP found that most system integrators are still running their old delivery model and adding AI in a few places, rather than rebuilding around it. Meanwhile a new wave of AI native players is changing how delivery actually gets done, and some are scaling fast. The message is simple. The big firms are moving slowly. The consultants inside them who move faster are pulling ahead. (McKinsey, 2026)
Second, the numbers got serious. Stanford’s 2026 AI Index reported that 88% of organisations now use AI, and that software teams are already 26% more productive with it. It also found a widening gap between what AI can do and how ready we are to manage it. That gap is not a problem to gloss over. That gap is the job. (Stanford HAI, 2026 AI Index)
Here is the pattern we saw again and again. Whether a consultant got value from these tools had almost nothing to do with the tools. It had everything to do with attitude. Same software, same login, two very different outcomes.
The consultants who won all did the same thing. They moved up a level. They let the agent take the work that never needed a human, and they spent the time they got back on the work that only a human can do.
Let me show you what that looked like across the five phases from the first post.
This one is live today. We saw it on a US project we delivered with our partner Wipro.
In the old world, gathering requirements meant weeks of workshops, interviews, and trawling through documents nobody had opened in years. It was slow, and most of it was clerical.
The agent changed the starting point. It read hundreds of documents on its own. Word files, PDFs, slide decks, spreadsheets, even images. From all of it, it produced around 400 requirements for the consultants to review.
Then it went further. It reverse engineered Java programs that had been changed and patched over ten years, and turned them into clear, readable requirements. For the first time, the consultants could see exactly what the old system did, and decide, line by line, what was worth moving to the cloud and what should be left behind.
That last part matters.
Deciding what to migrate is not clerical work. It is advisory work. It is the kind of judgement a customer pays a senior consultant for.
The smart consultants understood this instantly. They saw the 400 requirements as a head start, not a threat. They walked into the customer with that baseline on day one and used it to run a better conversation. What do you really need? What would add value to your business? What can we finally drop after ten years of carrying it?
They knew where their value had always been. It was never in reading every document and copying requirements into a spreadsheet. That is drudgery. Their value is in understanding those requirements, shaping them with the customer, and deciding what matters.
The agent handed them their time back so they could do the part only they can do.
Here is a sneak peak on how the Agent works:
Let me be clear about this one. It is not ready yet.
We are building the configuration agent now, and it will arrive in the coming months. But we can already see where it lands, because we have watched the same thing happen in software.
Two years ago, developers wrote every line of code by hand. Today agents write a huge share of it, and the developer describes the intent and reviews the result. The skill did not vanish. It moved up. The developer went from typing code to judging it.
ERP configuration is heading down the exact same road.
Instead of a consultant clicking through the same screens for days, they will describe the rule, and the agent will build the configuration. The consultant becomes the reviewer and the architect of intent.
We are already training it inside our own walls, and yes, it gets things wrong. In one case it produced a configuration that looked perfectly fine on the surface, but a functional expert spotted that it was wrong for how that business actually operated. So the consultant corrected it, and we taught the agent.
That is exactly how this is supposed to work, and we expect the same in the real world. The agent will be fast. The consultant will be right.
Here is the part consultants should hear clearly. When the agent takes the drudgery, the value of a good consultant does not fall. It rises. The work that actually moves a business forward is done by people who understand what the business needs. That work is not going away. It is becoming the whole job.
For years, testing required a team of people writing and maintaining test scripts, was normally the customers responsibility. It was slow, brittle, and mind numbing. Every change to the system required the scripts to be updated, and someone had to go and fix it by hand.
A while ago we told our partners they needed a test automation team just to keep up. Not any more.
Today the agent handles the execution. Is it fully autonomous? No. When a test fails, and it does fail, our team steps in, fixes it, and teaches the agent so it handles that case next time. But it is autonomous enough that we now tell our partners plainly: you do not need a separate automation team, leave the automation to us.
Over the last few months our confidence has grown so much that we are willing to put our own money behind that promise. It is not a line on a slide. It is a bet we are making.
So what happened to the people who used to write all those scripts?
The smart ones did not disappear. They moved up. They stopped writing scripts and started authoring scenarios. They took on the hard cases: the cross module flows, the exceptions, the situations where you have to understand the business to know what “correct” even means.
The agent runs the test. The consultant decides what is worth testing and reads what a failure is really telling them.
And some went further still.
Our partner Strategic Information Group (SIG) is doing something genuinely clever. Their CEO, Anise, and their functional expert, Keith, worked out that most test scenarios repeat across every single implementation. The same core flows turn up again and again, customer after customer, project after project.
So they are building a library. Hundreds of standard scenarios that are common across ERP rollouts, ready to go. For any new customer, they simply pick the scenarios that apply, run them through the automation module, and change the data. That is it.
Think about what that does. Work that used to take weeks now takes hours. Build a scenario once, reuse it forever. That library is a gold mine, and it grows more valuable with every project that adds to it.
This is exactly how innovative companies and consultants work. They refuse to do the same hard work twice. They turn one project’s effort into an asset that pays off on the next ten.
The value was never in running the test for the hundredth time. It is in building the thing that makes the hundredth test take minutes. The people who once feared the automation agent are now the authors of the most valuable asset in the practice.
That is not replacement. That is a promotion.
See the agent in action—here's a quick preview:
This is live, and it is where we learnt the most about people.
Our customer Constellation Cold Logistics (CCL), realized that a single user can make or break a go live. One person who is not ready can stall an entire rollout. That realization became our Driver’s License feature, a simple check that confirms each user can actually do their job before go live, not just sit through a slideshow.
You can read the full story here: Don’t go live until everyone is ready.
We also learnt that every user is different. Kent Klein, a Solution Architect at Columbusglobal.com said it best: "There are multiple ways to Rome, and I expect every user that makes it to Rome to pass the Drivers license test, regardless of which road they take".
That stuck with us.
So we trained our AI to adapt the training to the person, not just to the role. The trainer stopped delivering the same session for the fifth time and started leading the part that actually decides a go live: getting people to change how they work.
The result at CCL: 200+ users trained, across four languages, in a few months with a fraction of the normal work effort.
Here's how the Training Agent accelerates user training:
We are building this one now, and it might be the part we are most excited about.
Here is why.
Every ERP project has a memory problem. By go live, thousands of small decisions have been made, and no single person remembers all of them. Six months later, when a question comes in, the honest answer from even a very good consultant is often “let me go and find out.” That takes time. And sometimes the person who knew has already rolled off the project.
Our support agent never forgets. It understands the whole project, end to end. Ask it why something works the way it does, and it can tell you. It knows which requirements were agreed, and who agreed them. It knows how the system was configured, and why.
Ask it why automatic batch determination is switched off in the cold store, and it can answer: because the client chose manual batch selection for traceability during design, the logistics lead agreed it, and the sponsor approved it on that date. No archaeology. No guessing. No chasing three people who have all moved on.
It reaches into every phase. It knows how a scenario was tested and with what data, so when a pick confirmation that passed in testing starts failing in production, it can point out that the original test only ever used full pallets, while the new site runs mixed ones. It knows how the users were trained, so when operators keep asking how to do a partial putaway, it can see they were trained on full putaway only, and that the partial flow was added after their session.
That is not just an answer. That is the training gap, found and named.
These are exactly the questions that even senior consultants struggle to answer quickly. Not because they lack skill, but because no human can hold an entire project in their head. The agent can. It turns days of digging into a straight answer, and it does it at two in the morning when the warehouse is stuck.
But here is where it stops, and where the consultant becomes essential again.
The agent knows everything the project went through. It knows nothing about what the project never went through. When a user asks a question that needs real world experience the project never contained, the agent has nothing to draw on. The consultant has everything.
The auditors are coming next month and want to know how inventory is valued. A consultant who has sat through twenty audits knows what they will really push on, and how to get ahead of it. The business wants to switch a product line from make to stock to make to order halfway through the year. That was never in the design, and someone has to know what quietly breaks when you do it. The company is about to acquire another business and fold its warehouse into the system. No runbook covers that. Only scars do.
That is the line.
The agent gives you perfect memory of what happened. The consultant gives you judgement about what never has. One does not replace the other. Put them together and they beat either one alone, every time.
None of this was clean, and we are not going to pretend it was.
The agents got things wrong, so trust had to be earned rather than assumed. On one project the source data was such a mess that a person had to spend a week fixing it before any tool could use it, because the answer to bad data is still work. And the hardest problems stayed exactly as hard as they always were. Two directors who want opposite things is a negotiation, not a task. Getting a warehouse full of people to change how they work is the most human challenge there is. AI does not touch that.
We say this because the real worry behind the pushback was that we were overselling. Some of it, honestly, we were.
The consultant is not becoming a button pusher for a magic machine. The consultant is the person who catches the wrong config, fixes the data nobody else will touch, wins the argument in the room, and gets people to actually change. AI just clears the desk so there is time to do it.
It was never the consultant against the machine. It is the consultant who adopted against the one who did not, and they often sit two desks apart.
Every phase told the same story. The agent took the drudgery, and the smart consultant moved up. From typing requirements to shaping them. From clicking configuration to judging it. From writing test scripts to building a scenario library that pays off for years. From repeating training sessions to leading change.
The market says most firms are moving slowly. That is not a reason to relax. It is the opportunity. When the firm is slow, the person who moves becomes rare. And rare is the opposite of replaceable.
The consultants we watched adopt did not get replaced. They got a lot harder to replace.
Now turn the same lens on the company, because this is the part that should keep leaders awake.
This is not only about individuals. It is about whole organisations.
Strategic.com (SIG) is thriving, and it will keep thriving, because it is not leaving AI to one clever person in the corner. The whole team is adopting it. It is becoming part of how they work every day, on every project, for every customer. That is what compounds. That is what pulls a firm ahead and keeps it there.
The companies that do the same will keep pulling away. The companies that do not, no matter how big they are today, will be caught by this. It is a tsunami that has been building quietly for a while, and it is now close enough to see.
Size will not save anyone. Only adoption will.
The wave does not care how big you are. It only cares whether you moved.