If you are a large enterprise,
preparing to move a system you have used for twenty years
into a new SAP platform,
the most intuitive approach used to be:
hire a consulting firm.
Project managers arrive.
Engineers arrive.
Testers arrive.
Process consultants arrive.
A large IT project
could eventually involve dozens or even hundreds of people working long-term.
The consulting firms' business model was straightforward:
the more people needed,
the longer the project,
the more billable hours they could charge.
But by 2026,
a very interesting reversal is starting.
Enterprises originally hired consultants
to help them implement AI.
But once AI began to actually perform work,
one of the first costs companies reassessed was:
“Do we really need this many consultants?”
Bayer’s Six-Year SAP Transformation Now Supported by 30 AI Agents
German pharmaceutical company Bayer is undergoing a six-year major IT transformation.
One core task is reorganizing and deploying the company’s SAP system.
Such large enterprise system transformations
used to be major business for consulting firms.
Because it’s not just:
installing software.
It requires handling:
legacy code,
company processes,
data,
access rights,
testing,
systems in different countries,
and a large amount of customization.
But Jochen Kamp, who leads Bayer’s IT transformation, recently revealed
there are currently:
30 AI Agents
helping with programming and testing.
One of his goals for future deployment stages is to
require “significantly fewer” consultants.
The key point isn’t
that Bayer is about to fire all consultants tomorrow.
It’s that companies now have a chance to reevaluate
which tasks truly require an external consultant’s personal involvement.
Because the most expensive part wasn’t always the strategic decisions
The most costly parts of large IT projects
often aren’t the high-level strategy meetings.
The real drain on manpower
is the long list of execution tasks that follow.
Things like:
reading legacy code,
organizing specifications,
mapping system dependencies,
writing technical documentation,
modifying code,
creating test cases,
running tests,
finding bugs,
retesting,
preparing deployment documents.
Previously, each step might have needed different consultants,
engineers, and project staff investing significant time.
So much of consulting’s business was actually built on:
“This task requires many man-hours.”
But AI Agents disrupt this very unit of measurement—“man-hours.”
Even SAP itself predicts external consultant costs may decline
What makes this shift even more awkward is that
SAP, the enterprise software vendor,
has started discussing ways to reduce reliance on massive manual labor for deployment.
SAP recently talked publicly about a concept called:
deshoring
Previously, companies reduced IT project costs
through offshoring.
Meaning:
the work was still done by people,
but relocated to countries with lower wages.
For example,
work previously done in Europe or the US
was handed over to teams in India or other lower-cost regions.
But AI Agents bring a new question:
why only ask
“Which country’s people should do this work?”
when we should first ask:
“Does this work even need to be done entirely by people?”
SAP News Center recently cited an analysis of about 180 SAP conversion activities.
This analysis suggests that if AI Agents handle parts of specification, analysis, programming, and other tasks,
overall transformation costs might drop significantly.
Meanwhile,
Financial Times reported that SAP believes future systems embedded with more AI and automation
could reduce external consultant costs by up to 50%.
But a crucial caveat:
this is not yet a proven average outcome for all enterprises.
Capgemini CEO Aiman Ezzat called the 50% figure very optimistic,
saying they haven’t seen such widespread reductions yet.
However, he acknowledged:
SAP deployment is indeed being redesigned
to be faster
and cheaper.
So the key takeaway is not “definitely save 50%,”
but rather:
the consulting industry itself admits the same tasks may require fewer man-hours in the future.
Bristol Myers Squibb is already seeing some contracts disappear
A more direct case comes from pharmaceutical company Bristol Myers Squibb.
Greg Meyers, the company’s digital and technology head, said that previously, in cybersecurity,
BMS paid large sums to third-party firms
to have external personnel:
monitor systems,
detect anomalies,
and handle large volumes of daily security signals.
But now,
AI can perform much of this monitoring work.
The result is:
some of these external contracts are starting to vanish.
This change is even more critical than simply improving consultant efficiency.
If AI only reduced the time consultants needed from ten to five hours,
companies would still need consultants,
just faster ones.
But if the client’s AI system can complete entire segments of work independently,
the question becomes:
does this outsourcing contract still need to exist?
So companies are moving away from paying based on time spent
Another significant shift BMS is making
might truly transform the consulting business model.
The company is asking some external consultants to
reduce fees,
or avoid simply billing by
hourly rates.
Instead, they’re negotiating
fixed prices,
or even outcome-based fees.
The logic is straightforward.
Imagine a project that
once required ten consultants,
working six months.
Now AI helps consulting firms
accomplish it with five people in three months.
If the customer still pays based on the old manpower scale,
the efficiency gains created by AI
are almost entirely kept by the consulting firm.
Customers naturally ask:
“If AI makes you faster, why haven’t my bills decreased?”
So companies might be less willing to buy
“1,000 hours of consultant time.”
What they truly want is
“Was the problem ultimately solved?”
This aligns with what’s happening in India’s IT outsourcing industry
One of the IT service industry’s most important assets
has been:
people.
A client needed more work?
Deploy more people.
More engineers,
more analysts,
more consultants,
resulting in
more billable hours.
Many large service firms’ revenue
has long closely correlated with
number of people × hours worked × rate.
But now, with AI compressing some tasks,
“how many people were used and how many hours”
has become less important.
Clients want to know:
what was the outcome?
Did the system launch?
Did errors decrease?
Were costs lowered?
Did processes speed up?
If AI shortens a five-day analysis to one day,
clients won’t say,
“Please do it slowly over five days, we’ll pay the same.”
But the consulting industry is not disappearing
It’s easy to spin the story as
AI is going to kill consulting.
But current data doesn’t support such a simple conclusion.
Market research firm Source Global predicts that in 2026, companies may still spend about
$420 billion
on technology consulting services,
an 8% increase over 2025.
Consulting related to new technologies, including AI,
could account for roughly:
$140 billion.
In other words,
companies are using AI to reduce some consultant work,
while simultaneously needing more consultants in other areas.
This may sound contradictory,
but it is not.
Fewer execution tasks doesn’t mean fewer challenging problems
If AI can help with:
organizing data,
writing initial code versions,
building tests,
market research,
and monitoring cybersecurity signals,
then consulting work previously reliant on many man-hours
will naturally be reduced.
But enterprises still face many problems AI can’t easily handle themselves,
such as:
which process to change first?
what permissions should the AI system have?
why do employees resist using it?
how to resolve conflicting interests between departments?
how much risk can the board tolerate?
who is accountable if the transformation fails?
which legacy systems should truly be retired?
which work, although AI-capable, shouldn’t be automated?
These are not simple requests to
“generate an answer.”
They involve
internal company context,
power dynamics,
risks,
experience,
and crucially, someone must bear responsibility.
So what’s more likely to happen with consulting is not
disappearance,
but
compression of execution work and retention of high-value judgment work.
This is why understanding customers might become even more important
Home appliance company SharkNinja recently said they are reducing some consultant spend.
They use AI and technology partners
to complete more analysis and forecasting internally.
CEO Mark Barrocas pointed out a key reason:
internal personnel truly responsible for the business
know best about
their products,
customers,
promotions,
supply chain,
and what the company really wants to solve.
If AI can handle much of
data organization,
analysis,
and preliminary research,
companies will rethink whether
they need to hand over their data to external parties,
let outsiders spend months understanding their company,
only to bring back answers.
This is a real danger for consulting.
AI doesn’t just
make consultants faster,
it might also
enable clients to do well enough themselves.
But “can do it” isn’t the same as “doing it well”
We shouldn’t swing to the opposite extreme.
Just because a company installs ChatGPT,
Claude,
or a few AI Agents,
doesn’t mean it no longer needs external expertise.
Failed large ERP transformations can affect
finance,
inventory,
orders,
supply chains,
regulations,
even overall business operations.
Wrong cybersecurity judgments could have even higher costs.
Thus, companies should not look at AI and ask:
“Who can be cut?”
Instead, they should break down tasks:
which tasks are
high volume,
repetitive,
well-defined,
and verifiable?
Those can be prioritized for AI.
Which tasks need
major decisions,
handling exceptions,
cross-departmental coordination,
and high-risk judgment?
Those may still be worth paying experienced people for.
Consultants’ value will then no longer be:
“I have many people to do this for you.”
but rather:
“I know what must not be done wrong.”
AI may ultimately change what professional services really sell
Lawyers,
accountants,
consultants,
advertising agencies,
and software outsourcing companies
have long shared one common feature:
clients find it hard to precisely estimate how much time a task really takes.
So billing by hours
has been the easiest model to understand.
But AI is rapidly altering how long the same work requires.
What took three days yesterday
may take three hours today.
And perhaps three minutes next year.
As time becomes increasingly unstable,
professional service providers may eventually have to ask themselves:
If clients no longer want to pay based on “how long it took,” what value am I really providing?
It might be
outcomes,
judgments,
experience,
accountability,
risk-taking,
organizational change,
or
helping clients avoid very costly mistakes.
The most ironic part: consulting firms may be teaching clients how to rely on themselves less
This is the story’s most ironic twist.
Almost every large consulting firm now tells companies:
you must implement AI.
you must redesign processes.
you must integrate AI Agents into workflows.
you must increase productivity.
Once clients do these things,
they will naturally revisit all costs,
including
consulting fees.
So the biggest threat AI poses to consulting isn’t some model suddenly claiming:
“I can be McKinsey.”
It’s that companies slowly learn:
of the hundred tasks they used to outsource,
thirty AI can handle directly,
another thirty internal staff working with AI can complete,
and only the remaining difficult forty truly require external experts.
If this happens,
the most valuable consulting firm in the future
won’t be the one that brings the most people into the client’s company.
It will be
the one that solves the toughest problems with the fewest people.
Today, grow a little with AI.
Learn one AI skill every day.
Save a little time daily.
Improve your capabilities bit by bit.
SasaDaily, growing with you.
