The first phase of business AI was driven by access. Companies bought licences, opened pilot programmes and asked staff to experiment. The next phase is harder. It is about deciding where AI belongs, who checks its work and whether it improves a commercial result.
Several business leaders are now focused on that shift. Anthropic CEO Dario Amodei is pushing deeper enterprise partnerships, John Margerison and CEO of communications intelligence company XFactorAi, is building around human oversight and clearer decision-making.
Meanwhile IBM CEO Arvind Krishna is placing advanced models inside established security and business systems. Their approaches differ, but each points to the same conclusion. AI is moving from a stand-alone tool into the machinery of the company.
Here are five changes that will define what happens next.
1. Companies will stop rewarding AI use for its own sake
Some employers spent the past year encouraging staff to use AI as often as possible. A few tied usage to performance reviews or internal rankings. That approach is already being reconsidered.
The Financial Times reported that Amazon removed internal AI leaderboards after employees began automating low-value tasks simply to improve their scores. Other companies are also shifting attention from activity to results as token costs rise and weak outputs create more work.
This is a useful correction. A company does not benefit because an employee sends 100 prompts. It benefits when sales cycles shorten, customer complaints fall or a process takes half the time. AI use is not the outcome. It is one possible means of reaching it.
2. The best products will sit inside systems people already use
The AI market has spent years producing new tools. Businesses now have too many of them. The next winners will be companies that place useful AI inside existing software, services and processes.
Tata Consultancy Services recently announced a partnership with Anthropic aimed at scaling enterprise AI. IBM has also partnered with OpenAI to bring advanced models into security workflows. These deals matter because they combine new technical capability with existing customers, data, controls and distribution.
This supports one of XFactorAi CEO John Margerison’s central arguments. Adoption is more likely when AI becomes part of a familiar workflow rather than another platform that staff must learn, trust and remember to open. The technical model matters, but access to the point where work already happens matters just as much.
3. Acquisition will replace endless product imitation
Many AI start-ups have built impressive features without building a lasting route to market. Larger companies have the opposite problem. They have customers and infrastructure, but often move too slowly to build new capability themselves.
That gap is driving deals. Qualcomm agreed to buy AI software start-up Modular for nearly $4 billion, gaining software that can run models across different chips. C3.ai has also reportedly discussed a merger with Automation Anywhere.
This is not simply a race to own more AI. Buyers are looking for technology that fills a specific weakness, reduces dependence on a rival or can be sold quickly through an existing channel. Start-ups with no clear commercial position will struggle, even when their products are technically strong.
4. Human checks will become a design requirement
The promise of autonomous agents is attractive because it suggests software can complete whole tasks with little supervision. In many companies, that is still too risky.
Recent research into industrial AI adoption found that businesses often have more advanced experimental systems than they can safely put into production. The missing piece is reliable verification. Human review remains the main way to catch errors, protect confidential information and judge whether an answer fits the real situation.
That does not make AI less useful. It makes responsibility clearer. The best systems will show how they reached an answer, flag uncertainty and make approval easy. People should not be forced to choose between full automation and no automation at all.
5. AI strategy will become business strategy
IBM’s recent warning that companies are shifting spending from traditional software towards servers, chips and networking shows how far AI is changing corporate budgets. It is no longer a side project owned by an innovation team. It affects procurement, staffing, cyber security, suppliers and capital spending.
Leaders therefore need to make harder choices. Which models should the company depend on? What happens if a provider changes its terms or cuts access? Which work deserves automation, and which decisions must stay with people?
The companies that answer those questions well will not be the ones with the most pilots. They will be the ones that connect AI to a real business need, build it into everyday work and keep clear ownership of the final decision.
