The New Rules of Platform Strategy in the Age of Agentic AI

Alignment, not adoption, is what separates the leaders

Only 18% of companies say their AI, platform and business strategies are fully aligned. That single number explains most of the variation in what enterprises are getting back from artificial intelligence, because organisations inside that 18% compound returns while the rest fund pilots that never reach the core. Factor's research finds companies aligning all three dimensions achieve on average 2.2x higher revenue growth than peers.

The financial analysis behind that multiple is specific. Drawing on 602 public company statements, the average company grew revenue 6%. Companies whose platform strategy was well or fully aligned with their business strategy grew 10%, and those aligned with their AI strategy grew 13%. The same alignment shows up on the profit line as a 37% increase in operating profit, equivalent to roughly $1 billion in pre-tax earnings for the average enterprise.

Why a gap that size opens is not mysterious. When AI, platform and business strategies are set separately, an organisation buys the same capability twice, integrates it late, and measures it against three competing definitions of success. Alignment removes duplication, improves innovation velocity and concentrates effort on fewer, larger bets. Success with AI is an alignment problem, not a technology problem.

Most organisations have not made that move. Just 17% have embedded AI deeply into core business processes and more than 50% remain trapped in narrow pilots, because AI and platform decisions are made by different teams, funded through separate budgets and judged on different metrics. One global technology leader interviewed for the research warned that running AI as a separate initiative makes it very expensive, and that unless it is embedded in the overall strategy the organisation will never realise its value.

Most platforms are still pointed at productivity, not innovation

74% of organisations continue to focus on productivity as the main goal of their platforms, while only 60% view them as engines of innovation. That framing is self-fulfilling. A platform funded to make existing processes cheaper is scoped, staffed and measured for incremental efficiency, so incremental efficiency is what it returns: defended margin, no new capability.

Executives already know the framing has expired. 94% say the rise of agentic AI requires them to rethink their platform strategies, and more than half (57%) believe it demands significant change or complete reinvention. Broken down, 15% expect to overhaul their entire platform strategy, 42% expect a lot of changes and 37% expect some changes. Only 6% say they will not have to change at all.

The pressure is sharper than a rethink. Nearly one in five executives believe some of today's enterprise platforms will not survive the onslaught of AI within two years, because the design logic that made those platforms valuable, prescribed workflows, fixed hierarchies and rigid data models, no longer suits a world defined by autonomy and learning. As one platform executive put it, "The features that made platforms valuable are the ones AI is now outgrowing." A platform optimised to run existing processes more cheaply has no natural home for agents whose value lies in creating work that did not previously exist.

Two-thirds of organisations are improvising their platform strategy

Only 31% of companies say they have a formal, holistic strategy for how platforms are deployed across the organisation. Another 38% have one that is informal or piecemeal, 28% have concrete plans to develop one and 3% have nothing at all. The absence is not a documentation problem; it is the reason AI initiatives land in fractured environments where they cannot scale.

This is where the alignment gap becomes operational. Without a formal platform strategy there is no agreed inventory of which systems support which functions, no shared view of what is modern versus outdated, and no blueprint tying native vendor tools, third-party models and internally built agents together. It follows that 57% of leaders cite integration with existing systems as their top risk in scaling AI, ahead of the model, data and talent risks that attract more attention.

Fixing the foundation first pays back quickly. Mascoma Bank consolidated 66 fragmented systems, covering banking, CRM, loan servicing, digital channels, general ledger, insurance and wealth management, onto Salesforce Data Cloud and related platforms. Onboarding times fell from 45 minutes to just a few, a volume of paycheck protection loans equal to a full year was processed in only 13 days, and call centre hold times dropped by 98%.

Modernisation does not mean replacing everything, and leaders who treat it that way stall. It starts with a readiness assessment and a full inventory: map which platforms support finance, HR, supply chain and customer experience, then judge whether each is still fit for purpose. Brittle foundations turn AI into a liability rather than an advantage, as one ERP vendor leader put it while describing work to compress ERP implementation from six or seven years to under one.

The architectural choice: agents inside the platform, or across it

66% of organisations say they rely primarily on platform-native AI capabilities, while 32% are building platform-agnostic agents that span multiple systems. That two-to-one split is the most consequential architectural decision on the table, because it determines whether capability built this year is portable next year or trapped inside one vendor's assumptions about how work should flow.

The research argues the destination is hybrid, and sets out the structure that makes hybrid workable. A dedicated agent layer runs three tiers: utility agents that execute sets of basic tasks autonomously inside systems, super agents that understand user intention and mobilise the right utility agents to achieve a goal, and orchestrator agents that assign tasks and coordinate across multiple super agents. Utility agents typically live inside specific platforms; super and orchestrator agents can live outside them.

A financial close makes the model concrete. Utility agents retrieve balances and validate journal entries inside SAP and Workday to stay compliant. Super agents interpret the wider goal, such as finalising fourth-quarter results, and manage dependencies like locking payroll data before journals post. Orchestrator agents track progress and preserve auditability through enterprise control planes.

Nearly six in 10 organisations are already planning structural adjustments to accommodate AI. The practical work for CIOs is building abstraction layers that free agents from having to understand the intricacies of multiple systems, improving data quality, and establishing integration patterns that support real-time, trustworthy execution. Factor's research on how far Australian organisations will let systems act autonomously is the companion read on where those boundaries are being drawn.

Work divides three ways, and the ratio changes by function

Beneath the survey sits a task-level analysis that gives the platform-human-agent question a measurable shape. Factor evaluated 332 intermediate-level tasks representing the activities of the workforce, classified them using O*NET Online's taxonomy of work activities, weighted them by US Bureau of Labor Statistics data on employment and hours worked, and estimated what share of work time in each function can be handled by platforms, agents or humans.

The spread is wide enough to change where an organisation should start. Customer service is the most agent-intensive at 58%, with 9% platform-intensive and 33% human-only. IT follows at 53%, 21% and 26%, then marketing at 52%, 9% and 39%. Sales and operations both sit at 47% agent-intensive, and HR at 42%, with 42% of HR work time remaining human-only.

The two ends of the range matter most. Finance is the most balanced, at 40% agent-intensive, 26% platform-intensive and 34% human-only, which is why it rewards a three-way design rather than an automation push. Legal is the opposite: 27% agent-intensive and 10% platform-intensive against 63% human-only. That human-only share is the number to plan against, because it varies by more than a factor of two across functions, and a single automation target applied everywhere will overshoot in some and leave value untouched in others.

Agents arrive function by function, not enterprise-wide

Executives do not expect agentic AI to hit every process at the same speed, and the variation is wide enough to drive sequencing decisions. Sales and customer management leads in AI augmentation, with 50% of companies using AI for chatbots, personalisation and enablement. Product and innovation is close behind at 48%, with AI embedded in research and development, design and testing.

Expectations of outright transformation cluster elsewhere. IT and digital stands out, with 34% of organisations expecting agents to be transformational, while marketing and branding, product innovation, and HR and talent each show 25% to 26% expecting agent-first execution models. In those functions the structure of work is being redefined, with agents orchestrating tasks end to end rather than assisting with steps inside a human-run process.

The cautious end is equally instructive. Risk and compliance leads in supplementary use, with 41% using AI to enhance functionality while leaving core workflows intact, a rational response to regulatory sensitivity. Strategy and planning follows, with 32% using AI as an advisor rather than an executor. Finance shows the most resistance, with 15% of companies reporting no AI impact at all.

Where deployment has happened, the returns are specific rather than promissory. Lenovo used Adobe Experience Platform and Microsoft Copilot across marketing, customer service and internal workflows, delivering $11 million in efficiency savings and a 12.5% boost in click-through rates. Western Sugar now processes about 40,000 invoices a year with no human intervention until approval, cutting processing time 25%, and US AutoForce cut reconciliation time by 80%, saving more than 30 hours a month. Zurich Insurance's AI-powered CRM reduces service times by over 70%, and Factor's work on how agentic systems are reshaping commercial motion tracks that shift through the deal cycle.

The adoption barrier is training and trust, not resistance

64% of companies cite employee resistance as the biggest barrier to scaling AI, and that finding is routinely misread. The next two barriers explain the first: insufficient training programmes at 51% and limited training budgets at 47%. When people are not given the skills, understanding or support they need, hesitation is not resistance, it is a rational response to being asked to change how they work without being equipped to do it.

There is a trust dimension as well as a skills one, and it carries a measurable price. Nearly three-quarters of companies had to pause at least one project in 2024 due to AI-related risk, according to earlier Factor research, which is what black-box logic and unresolved governance look like on a delivery schedule. The reverse also holds: companies can expect a 25% increase on average in customer loyalty and satisfaction from offering responsible AI-enabled products and services. Factor's Australian risk research covers how boards frame that exposure.

Factor's own Marketing and Communications team ran the change on itself, redesigning more than 2,000 roles to reflect a new model of human-agent collaboration. It clarified who does what between people and agents, created new metrics such as agent output effectiveness and intervention frequency, and introduced routines to govern agent contributions. Campaign steps fell from 135 to 85, getting the average campaign to market 25% to 35% faster.

Structure has to move with skills. As automation reduces the need for junior execution, the classic pyramid gives way to a diamond, where mid-level roles such as product owners, domain architects and solution integrators become the centre of gravity, supported by the roles the research calls trainers, explainers and sustainers. Without investment in that middle tier, transformation stalls whatever the technology does.

What this means for Australian organisations

Read the sample honestly before applying the numbers. This research surveyed 1,031 C-suite and senior executives at companies based in 12 countries and across 10 industries, at enterprises with revenues greater than $500 million, supported by 20 structured interviews. Australia is one of those 12 countries and accounts for 5% of respondents, alongside the United States at 29% and France and Germany at 10% each. These are global benchmarks that include Australian executives, not an Australian survey.

That distinction matters less than it might, because the vendor landscape the research describes is close to identical here. Australian enterprises have standardised heavily on the same core platforms across finance, HR, service, supply chain and customer operations, and every one of those vendors is now shipping an agent layer. The 66% relying on platform-native capability versus 32% building platform-agnostic agents is a choice Australian CIOs are making right now, usually without a formal platform strategy to make it against.

Australia's market structure sharpens two findings. Concentration in banking, insurance, telecommunications and retail means a small number of large incumbents run very similar platform estates, so platform-native agent capability delivers parity rather than advantage; differentiation has to come from the processes the Platform-Agent Impact Map places in the high-value, high-differentiation quadrant. And the regulated profile of those same sectors makes the 41% supplementary posture in risk and compliance the likely Australian default, defensible so long as it is a decision rather than a drift.

Scale cuts the other way. Australian enterprises are large by domestic standards and mid-sized by the standards of this sample, which means smaller architecture teams carrying the same integration burden that 57% of leaders name as their top scaling risk. That argues for a narrow agenda executed properly rather than a portfolio of pilots that each fight the same fractured foundation.

The talent finding lands hardest locally. A market this size has a thin supply of the orchestration, integration and AI-fluency skills the research says are now central, so the 51% citing insufficient training programmes and 47% citing limited training budgets translate into an Australian capability constraint that hiring alone cannot solve. Building trainers, explainers and sustainers internally is slower than recruiting them, and here it is usually the only path.

What Australian technology leaders should do next

Test alignment before approving the next tranche of AI funding. If the AI roadmap, the platform roadmap and the growth plan are not answering the same question, the 2.2x revenue growth advantage in this research is not available at any level of spend, and only 18% of companies currently pass that test.

Write the platform strategy down and make it holistic. 31% of companies have a formal, comprehensive platform strategy and the rest are improvising, which is what leaves AI initiatives fighting environments they cannot scale in. Inventory every platform by function, assess which are modular, cloud-native and AI-capable, and name what is outdated or missing.

Decide the agent boundary deliberately rather than by vendor default. Follow the 32% building platform-agnostic capability for the super and orchestrator layers that must span systems, while letting utility agents live natively inside the platforms they operate in. Capability locked inside one platform's assumptions is capability you may have to rebuild once the agentic layer settles.

Move part of the platform mandate from productivity to innovation and fund it as such. 74% pointing platforms at productivity is the global default, and defaults do not produce outperformance; use the Platform-Agent Impact Map to pick the few high-value, high-differentiation processes worth reimagining and standardise the rest. Sequence with the task evidence, starting where the agent-intensive share is highest, such as the 58% in customer service.

Fund training before diagnosing resistance. 64% of companies blame employee resistance while 51% report insufficient training programmes and 47% report limited training budgets, and the second pair causes the first. Clarify who does what between people, platforms and agents for one critical workflow, then scale the model. See also Factor's research on customer service under pressure in Australia, explore the full research library, or join a Factor CIO or CTO event.