Q4 2025 Consulting Trends: AI Agents & Pricing Shift

From the market, job and consulting industry developments in Q4 2025, we observed the consulting practice evolve from a strategy-first engagement to an implementation-focused service.

Broadly, we observed 5 trends:

1) Consulting Trend #1: Hardware and Privacy-aware AI Integration
2) Consulting Trend #2: Strategy advisory to AI-embedded execution models
3) Consulting Trend #3: Outcome-based pricing is the baseline client expectations
4) Consulting Trend #4: AI Agents in Governance and Compliance
5) Consulting Trend #5: Skills in AI Orchestration

1) Consulting Trend #1: Hardware and Privacy-aware AI Integration

a) Hardware-Aware AI

The biggest roadblock to scaling AI is energy costs.

Nuclear Energy as a source of training AI models has pushed technology giants Amazon, Microsoft, Meta, and Alphabet ^[1]to invest in massive Nuclear infrastructure, but for most enterprises with shallower pockets, adopting hardware-aware AI models is the only way to overcome the energy shortage.

Even Outsourcing Giant TCS, recognizing its limited lifeline to climb out of delivery disruptions, has partnered with AMD to co-develop an AI/GenAI solution exclusively for AMD's Instinct GPUs, EPYC CPUs, ROCm software, and rack-scale systems ^[2]. Even the software service's architecture is designed and optimized for specific AI hardware to ensure operability throughout the partnership.

b) Privacy and Local LLM

A reason enterprises and customers hesitated to embrace AI outside the mandate of technology giants is the uncertainty about how AI will be adopted.

From skimming entire emails after a user missed one opt-out button, to operating systems consuming entire documents without explicit permissions, customers had lost trust in Technology giants' openness about using PII data in early iterations of GenAI.

Enterprises operating in industries with strong regulatory oversight (healthcare, finance, defense, and legal) had to invest in dedicated AI hardware to implement on-premises AI to comply with laws and regulations such as GDPR, HIPAA, the EU AI Act, and CCPA.

As of Q4 2025, around 55% of enterprise AI inference now runs in private AI stacks and on-premises, without relying on cloud service providers ^[3]. The demand for local AI has spurred new open Models in Robotics from NVIDIA, for multiple purposes, including simulations for Physical AI (Cosmos), humanoid (GR00T), Policy Evaluation in Robotics (Isaac Lab-Arena), and edge computing (Jetson T4000)

As enterprises and, eventually, tech-savvy consultants adopt the Local LLM architecture, the pressure on cloud-based AI service providers, including Gemini and OpenAI, will be felt.

When the workforce adopts a two-tier architecture, the surviving LLMs would be the ones offering niche models for diverse use cases.

c) Warehouse and Operations: Robotics and AI Couldn’t Replace Humans

Even after the rapid investment in Physical AI, there are use cases where the moving parts of the operations, purely due to their scale, add to the implementation complexity.

A notable failure was Amazon's Blue Jay project, which went into production in October 2025. The robots couldn't adapt to smaller warehouses, disrupting the project's objective: "To deliver same-day packages" from smaller spaces closer to the customer.

But the technology behind the project wasn't entirely lost. They found a new avatar in Orbital ^[4], a modular configuration for warehouse management. Right now, Amazon is stuck with DeepFleet to optimize robotic paths for warehouse management.

Project Eluna, an agentic model where the key operations person is central to analyzing real-time data, has grounded Amazon to the realities of rigidly following AI transformation in operations.

Amazon's aggressive cost-cutting dreams where the number of humans in the human-robot ecosystem is optimized for cost is now at a turning point.

2) Consulting Trend #2: Strategy advisory to AI-embedded execution models

McKinsey and AWS launched the Amazon McKinsey Group (AMG), an integrated strategy-to-implementation operating model.

Strategic Partnership Between Consulting and Technology Giants

AWS, a stable revenue source for Amazon from 2020 to 2025, faced a new challenger in Google Cloud.

With Gemini integrated into Google's cloud offering, AMG is Amazon's early strategic reaction to the 29% market share in 2025, the first time AWS share fell below 30%.

While the market matures and enterprise AI adoption crosses the critical threshold, consulting companies have mimicked NVIDIA's circular deal strategy: foundational model companies pay for the hardware, and NVIDIA buys their models.

McKinsey has adopted a similar playbook by forging strategic partnerships with Amazon to lock in AWS and similar proprietary platforms across its entire operations.

With clients increasingly seeking outcome-based consulting services, the Big Cs have no option but to collaborate with technology giants and transform their consulting solutions into measurable, high-impact interactions.

Information Asymmetry of Consulting – Disrupted

With close to 10% of the workforce, mostly Junior Consultants in the US and developed economies, facing layoffs, advanced AI tools like Claude and GPT-5 have taken over the role of junior consultants.

Traditionally, what it took junior consultants, weeks can be done by these AI's in just minutes with the right training.

The 'information asymmetry' buried in a large consulting firm's internal databases has limited value in guiding clients' strategy.

Pure strategy formulation roles are dead.

But with each disruption, a new niche evolves.

Fastest Growing: Strategy Expertise

Strategy execution, making sense of the strategy, and orchestrating AI agents with automation, compliance, and scenario planning, are the fastest-growing strategy expertise in Consulting.

With a brand-new technology like AI agents, the execution is not straightforward.

The consultant must partner with technology firms, collaborate and co-design workflows, upskill the team, and create a governance framework with strong guardrails against data leaks and unethical AI actions.

Transitioning from a Junior Consultant to an Orchestrator role is a steep learning curve. The Consulting companies have the leverage to keep the salaries stagnant, which explains the 3-year stagnation in MBA salaries in Consulting.

Most mid-tier MBA programs experienced a hiring dip in Consulting, from 40%+ highs in 2023 to the low 30%- 20% in 2025.

For the first time, AI-related job postings have overshadowed consultant postings ^[5]

Bad News: Junior Consultants

The planned layoffs in large consulting and accounting firms paint a grim job market^[6]:

•    PwC plans to cut U.S. campus hiring by 30% until 2028
•    McKinsey plans a cut of 10% in non-client-facing roles (approx. 4,000) until 2027
•    PwC: close to 1% of the global headcount in 2025 (mostly audit/assurance, legacy technology, HR, and marketing)
•    Accenture: Close to 1.5% in 2025, mostly laying off roles that have no relevance in a post-AI world.

Indirect Pressure on Junior Consultants

Junior Consultants face indirect pressure to acquire AI and domain expertise to grow, while mid- and senior-level consultants also view junior consultants as a threat, as roles earmarked for research and due diligence are now delivered by AI.

Conspiracy Against Junior Consultants

Bloomberg News reported that ex-consultants from the big consulting companies were training AI models (code-named Project Argentum) with Gemini to take on most of the junior consultants' roles, including building PowerPoint slides, defining problem-solving logic, cleaning and analyzing data, and conducting market analyses.

While there is no follow-up on progress, Agentic AI has accelerated adoption, with McKinsey creating internal AI agents to disrupt the pyramid model of delivering consulting service.

3) Consulting Trend #3: Outcome-based pricing is the baseline client expectations

Inside the intricacies of the negotiation for outcome-based pricing is one fact: only 25–30% of large consulting company's revenue are driven by outcome-based consulting models.

Fixed-income fees constitute 45-47% of total revenue.

Large Consulting Companies: Embracing Fixed-Income Model

The integration of AI has disrupted this guaranteed income, with estimates indicating annual growth of 30-35% in performance-based fees. ^[7]

If we use a fixed fee as the proxy for revenue growth in large consulting companies, the growth is only 4-7%.

Even by conservative estimates, outcome-based pricing will be the norm in 2029.

Large consulting companies are proactively embracing this reality by establishing frameworks for pre-project data validation and creating AI-driven models to evaluate causal inferences and demonstrate that Consulting had an impact on the client's outcomes.

Mid-Tier Consulting Companies: Execution Roles

Mid-Tier Consulting companies don't have the leverage of long-term relationships and dependencies on their service.

Q4 2025 Consulting trends indicate they would be forced to adopt the role of the primary 'execution' team, bringing in a cohort of consultants and technologists under a new hybrid delivery model.

Accounting Pivot

Accounting firms had to pivot aggressively, choosing two different strategies. EY focused on Agentic AI in advisory roles from compliance to due diligence ^[8] while PwC is integrating AI across the workflow with AI studios^[9]

4) Consulting Trend #4: AI Agents in Governance and Compliance

While most consulting companies are breaking down the pre-project audit, KPMG has announced a strategic relationship with Uniphore to build small language models ^[10] to serve regulated industries such as Oil and Gas, healthcare, telecommunications, customer service, and financial services, where compliance is integral to the consulting project.

With 60% enterprises lacking a formal governance framework for controlling AI and AI agents, new demand will emerge that merges traditional cybersecurity with a kill switch and other guardrails to monitor autonomy thresholds, a measure of an AI agent's independent decision-making.

Role of regulation

With President Trump adopting a light-touch approach to regulating AI, technology leaders and global citizens will keenly await the EU AI Act ^[11], set to come into effect by August 2026. Among the broad technical frameworks, human oversight and quality controls, the 7% global turnover penalty, up to €35M, could create a new regulatory compliance role in consulting.

Even the liberal US states have adopted a light-touch approach to avoid curbing innovation with the Colorado AI Act (effective June 30, 2026)^[12] , which evaluates the impact of AI on discriminatory behavior in decision-making, while California SB 53 ^[13]mirrors current incident-reporting frameworks in cybersecurity.

The voluntary regulation of AI companies will end by mid 2026.

The compliance consultants will closely monitor the penalties and remediation framework in Agentic AI.

5) Consulting Trend #5: Skills in AI Orchestration

From AI literacy to model training, Q4 2025 saw the emergence of demand for AI orchestration skills.

Unlike pure technology roles, in Consulting, skills are valued for building AI orchestration frameworks (including LangChain and AutoGen).

The most lucrative roles would be for orchestrators with strong domain skills in technology and finance.

Healthcare and manufacturing are where multiple departments, regulations (cross-border), cross-functional teams, processes, humans, machines, and AI intermingle.

It is in these complex interactions that the new generation of consultants would shine.

Roles like AI Agent Orchestration Specialist^[14], Intelligence Engineer, AI Operations/Platform Engineer, and Agentic AI Manager will emerge in our vocabulary by mid 2026.

The most in-demand AI consulting skills for 2026 require:

1)    Translating ML models into feasible implementation timelines
2)    Fine-Tuning the model for the enterprise’s use case
3)    Strategy for scaling the models
4)    Evaluating advanced models for special use cases
5)    Data planning and design to meet regulatory and security requirements

References

[1]Nuclear Powering AI