Corporate AI Training Healthcare Karachi, Pakistan 2025

220 staff trained. 7.2 hours saved per employee per week. 89% tool adoption in 4 weeks.

The Aga Khan University Hospital engaged Densight Labs to design and deliver a department-specific AI training programme for non-clinical administrative and operations staff. The goal: measurable productivity gains, without disrupting clinical workflows or requiring IT infrastructure changes.

220Staff trained across 5 departments
7.2 hrsSaved per employee per week (Week 4)
89%Tool adoption rate by Week 4
64%Faster administrative report generation
3.1×Estimated Q1 ROI from hours recovered

40–60% of working hours spent on tasks AI could handle — but no tools, no policy, and no plan

Administrative teams across Aga Khan University Hospital were spending nearly half their working day on tasks that generative AI could compress or automate: drafting reports, writing correspondence, processing procurement documents, searching databases, and formatting HR documentation.

The hospital had no existing AI policy, no enterprise tool licences, and significant staff resistance rooted in fear of job displacement. Leadership wanted measurable results — not a pilot programme that lived in a slide deck. Densight Labs was brought in to deliver training that worked at department level and produced documented outcomes within 4 weeks.


Department-specific AI training built on the ADAPT Framework — not a generic AI workshop

Before Training

  • AI Readiness Survey across all 220 participants
  • Workflow mapping by department — identified top 3 use cases per team
  • Addressed job displacement fears with data during brief pre-session

Day 1 — Foundation

  • How LLMs work (no jargon, no hype)
  • What Claude, Copilot, and Perplexity can and cannot do
  • How to write effective prompts for administrative work
  • Live demos using actual hospital document types

Day 2 — Live Workflows

  • Split by department — no generic exercises
  • Finance: AI-assisted report drafting in Claude
  • HR: offer letter and JD generation prompts
  • Medical Records: Copilot document summarisation
  • Every participant left with 2+ working AI workflows

After Day 2, Densight Labs ran a 4-week follow-up: dedicated WhatsApp support channel, bi-weekly adoption dashboards for department leads, and a Week 4 measurement audit. This is the Track phase of the ADAPT Framework — we don't call a training done until we've measured what changed.

Medical Records

48 staff. Primary use cases: document summarisation, correspondence drafting, patient record formatting. Tool: Microsoft Copilot + Claude.

Finance & Billing

38 staff. Primary use cases: monthly report drafting, reconciliation write-ups, budget variance explanations. Tool: Claude.

Human Resources

34 staff. Primary use cases: offer letters, job descriptions, policy document drafting. Tool: Claude + Notion AI.

Procurement

48 staff. Primary use cases: RFQ generation, supplier comparison write-ups, approval memos. Tool: Claude.

Patient Services

52 staff. Primary use cases: patient correspondence, FAQ responses, appointment coordination emails. Tool: Claude + Copilot.

Tools Introduced

Claude (Anthropic)
Microsoft Copilot
Notion AI
Perplexity Pro

7.2 hours recovered per employee. Every week. Across 220 people.

7.2hrs
Average weekly time saved per employee (Week 4)
89%
Staff using 2+ AI tools daily by Week 4
64%
Reduction in time to produce administrative reports
3×
Faster clinical correspondence drafting
78
Post-training NPS (industry avg for L&D: 32)
3.1×
Estimated Q1 ROI from productivity hours recovered

Before Training

  • Monthly finance reports: avg 2.4 hrs to draft
  • Procurement RFQs written from scratch each time
  • HR offer letters: 35–45 min per letter
  • Patient correspondence: dictation + manual typing
  • Data lookups: manual search across multiple systems
  • No AI tool licences across any department

After Training (Week 4)

  • Finance reports: avg 52 min — 64% reduction
  • Procurement RFQs: AI-drafted in 8 min from prompt templates
  • HR offer letters: 4–6 min with Claude (90% reduction)
  • Patient correspondence: AI-drafted, human-reviewed — 3× faster
  • Perplexity Pro cutting research time by 70%
  • 89% of staff using 2+ AI tools daily

From readiness survey to Week 4 measurement audit — every step documented

Week −1 (Pre-Training)

AI Readiness Assessment & Use Case Mapping

All 220 participants surveyed. Workflows mapped by department. Highest-value AI use cases identified per team. This is the Assess + Diagnose phases of the ADAPT Framework — no generic training agenda until we know the exact pain points.

Day 1

Foundation Workshop — How AI Works & What It's For

Conceptual grounding in LLMs — no jargon, no hype. What Claude, Copilot, and Perplexity can and cannot do. How to write effective prompts for administrative work. Job displacement addressed directly with data. Delivered by Numan Ahmad, Founder of Densight Labs.

Day 2

Live Tools — Every Participant Builds Real Workflows

Department-split sessions. Finance: AI report drafting. HR: offer letter and JD generation. Medical Records: document summarisation with Copilot. Procurement: RFQ prompt templates. Every participant left with 2+ working AI workflows configured in their own accounts — not demos, live tools.

Weeks 1–4

Follow-Up, Adoption Monitoring & Week 4 Audit

Dedicated WhatsApp support channel. Weekly office hours. Department leads received bi-weekly adoption dashboards. Week 4 measurement audit compared actual time savings against pre-training baseline. Results documented and shared with hospital leadership.

"The shift we saw by Week 4 wasn't just efficiency — it was confidence. Staff who were afraid of AI on Day 1 were building their own prompt libraries by Week 3. That's what proper AI implementation looks like."

— Numan Ahmad, Founder & CEO, Densight Labs

Want results like this in your organisation?

This is what corporate AI training looks like when it's built around your team's actual workflows — not a generic AI overview. Book a discovery call and we'll map the training to your context.