Copilot and Agent Adoption in the Big Four Professional Services Firms

Mo Khalifa, an AI Workforce Specialist at Microsoft covering the Big Four professional services firms, returns to the podcast after a spell at HP Poly and a previous stint in the Teams Rooms business. He covers how these highly regulated, multi-member-firm organisations are approaching Copilot rollo

Copilot and Agent Adoption in the Big Four Professional Services Firms

Prefer audio?

Who should listen: M365 service owners, change and adoption leads, and IT decision-makers planning or scaling a Copilot and agent rollout in large, regulated or multi-entity organisations.

Guest: Mo Khalifa — AI Workforce Specialist, Microsoft

Mo Khalifa, an AI Workforce Specialist at Microsoft covering the Big Four professional services firms, returns to the podcast after a spell at HP Poly and a previous stint in the Teams Rooms business. He covers how these highly regulated, multi-member-firm organisations are approaching Copilot rollout, agent-to-agent workflows, data sovereignty and the practical mechanics of adoption.

Many thanks to AVI-SPL, sponsor of this episode.

Key insights

  • Copilot adoption should be treated as habit formation, not a training event: budget roughly 28 days of daily use before it becomes second nature, and don't expect users to 'learn it' in a fortnight. ▶ 10:43
  • Sequence the journey in three stages — personal productivity first (morning email triage, action items where you're mentioned), then a team-shared agent, then agents wired into business processes. ▶ 11:14
  • Khalifa built his own agent in Copilot Studio (formerly Agent Builder) over a central repository of customer material — EA agreements, past conversations — so he can ask which customer said what before a call rather than searching manually. ▶ 12:15
  • Microsoft's own onboarding is now agent-driven: a scanned barcode triggered an employee agent that ordered the laptop, set up MFA and explained holiday policy — an example of agents chaining to procurement or ServiceNow ticketing rather than working in isolation. ▶ 13:17
  • A Big Four customer's framing that stuck: as fleets of agents start doing whole swathes of work, today's IT teams effectively become the HR teams for agents — which is where Agent 365 as a management and control layer fits. ▶ 14:48
  • Data classification with Purview is the number one prerequisite. Where labelling is incomplete, apply a zero-trust default — assume compromise, label everything highly confidential and make users request access — which organically builds clean data over time. ▶ 22:55
  • Be explicit about licensing: Copilot Chat is free with M365, Windows and Edge and only reasons over what's on screen plus the web; M365 Copilot reasons over enterprise data as well. Many customers assume they already have the paid product. ▶ 26:31
  • Start rollout where the pain is obvious — Teams meeting recaps are the single most-used entry point, then Outlook meeting prep — and rewrite the generic Microsoft prompt libraries into role-specific examples (a day in the life of an auditor, a tax specialist). ▶ 29:04
  • Work IQ addresses a specific Big Four ask: organisational memory for firm-specific acronyms, project names and regional language, so prompts no longer need re-priming with context every time. ▶ 19:53

Insights summarised by AI from the episode transcript, reviewed by the Empowering.Cloud team.

Listen: Apple Podcasts · Spotify · Other platforms