AI for HR, finance, marketing, ops

AI for Marketing and Sales: A 90-Day Pilot Framework With Scoped Budget Tiers

· By AIHQ Team

What HR and L&D leaders actually control in a marketing and sales AI pilot

Two of your colleagues want budget for AI next quarter. The marketing lead wants copy support for campaign volume. The sales lead wants better follow-up on inbound enquiries. Both requests arrive as tool subscriptions, and neither comes with a baseline, a measurement method or a defined end date.

That is the position HR and L&D leaders are usually in with AI for marketing and sales: not choosing the tool, but deciding whether a pilot is structured well enough to be worth funding. Training sits inside that decision whether or not anyone names it, because a licence nobody uses is a cost, not a capability.

A 90-day frame with two locked workflows, a baseline recorded before launch, and three clearly priced budget tiers gives you something you can approve, review and stop.

Lock two workflows before you discuss any tool

A pilot that touches eight workflows tells you nothing. A pilot that touches two tells you whether the team can adopt anything at all.

Pick workflows that are high frequency, low regulatory risk, and already measured in some way by the department. Two that consistently qualify:

  • Campaign first drafts. Brief-to-draft cycle time for email, landing page or social copy. Marketing managers already track how many assets ship per month.
  • Inbound follow-up. First-response time and follow-up completion rate for enquiries that arrive through web forms or event lists. Sales ops usually has this in the CRM.

Both have a before-and-after number that does not depend on anyone's opinion. Both also have a visible quality dimension, so the pilot cannot be judged on speed alone.

If your organisation wants a wider net of candidate workflows first, an AI use-case discovery workshop is the cheaper way to sort real pain points from enthusiasm. But once the pilot starts, two workflows is the ceiling.

Set the baseline in week one, not week twelve

Baselines are the step most pilots skip, and the reason most pilots cannot be defended at review. Ask each department head for four numbers before a single licence is activated:

Metric Marketing owner Sales owner
Volume Assets produced per month Enquiries handled per month
Cycle time Brief to approved draft (hours) Enquiry to first response (hours)
Quality Revision rounds per asset Follow-up completion rate (%)
Review load Manager hours spent reviewing drafts Manager hours spent checking follow-ups

The fourth row matters most for L&D. If output volume rises but senior review hours rise with it, nothing has been gained. Record these numbers in the same document as the budget so nobody redefines success later.

Week-by-week rollout with named owners

This is the timeline to put in front of your approval forum. Owners are roles, not individuals, so the plan survives staff changes.

Weeks 1–2 — Baseline and scope. Owner: L&D lead with marketing and sales ops. Collect the four metrics per workflow, confirm the two workflows, and agree the review date. Deliverable: one-page pilot charter signed by both department heads.

Weeks 3–4 — Access and guardrails. Owner: IT with risk or compliance. Provision accounts, agree what customer and prospect data may enter a tool, and issue a one-page usage note covering confidential information, client names and draft output review. This is also the point to check your obligations under Malaysia's PDPA if personal data is involved. A responsible AI and governance session at this stage is cheaper than retrofitting policy after an incident.

Weeks 5–6 — Role-based enablement. Owner: L&D with a trainer or internal champion. Two half-day sessions per team, built around the two locked workflows rather than generic prompting. Marketing writes real campaign briefs. Sales works real enquiry threads. Expect the output quality gap to be widest here, which is why the prompt engineering skill matrix is worth using to set expectations by job family.

Weeks 7–10 — Supervised use. Owner: department heads. Teams use the tools in live work with a named reviewer for every output that leaves the department. Weekly 20-minute stand-up captures what worked, what needed heavy editing, and what should be dropped.

Weeks 11–12 — Measure and decide. Owner: L&D lead with finance. Re-run the four metrics, compare against baseline, and produce a one-page finding: continue, expand, or stop. Include a cost-per-workflow figure so the next funding conversation starts from a number.

Weeks 13 — Governance handover. Owner: risk or compliance. Convert what the pilot taught into a short standing policy for ongoing use, including which tools are sanctioned and who reviews output.

Three budget tiers and what each one realistically buys

Budget conversations go sideways when the tier is unclear. These three tiers are scoped to a pilot covering roughly 15–30 users across the two workflows, over 90 days.

Criteria Tier 1: Tooling only Tier 2: Tooling + training Tier 3: Tooling + training + implementation support
Team licences for pilot users Yes Yes Yes
Baseline design and measurement Department does it themselves Light support Designed and tracked with you
Role-based training None Two sessions per team Two sessions plus workflow coaching
Prompt and workflow library None Basic starter set Built for your two workflows
Governance note and usage policy Department drafts it Reviewed by L&D Drafted with risk or compliance
Integration with CRM or CMS Not included Not included Scoped separately
Realistic 90-day outcome Usage data and anecdote Usage plus measured workflow improvement Measured improvement plus a reusable workflow asset
HRD Corp claimability Not applicable to licences Training portion may be structured to be claimable Training and workshop elements may be structured to be claimable

Two honest caveats. First, no tier guarantees a productivity outcome — results depend on adoption, review discipline, data quality and how the work is measured. Second, HRD Corp claimability depends on client eligibility, grant approval and HRD Corp submission requirements, so treat the training line as potentially claimable rather than claimable by default. The HRDC claim rejection patterns are worth reading before you build the training line into the budget.

Who should pick which tier

Pick Tier 1 if your marketing and sales teams are already confident AI users and the pilot is really a measurement exercise. Expect to spend internal time on baseline design and training, and be honest that adoption will be uneven.

Pick Tier 2 if the teams are mixed — some enthusiasts, many cautious users — and the goal is to see whether structured enablement changes anything. This is the most common starting point for organisations that have already tried ad-hoc tool access.

Pick Tier 3 if the workflows touch customer data, if the two workflows need to connect to your CRM or CMS, or if leadership wants a repeatable process rather than a one-off experiment. Tier 3 is also the right answer when the pilot is expected to become a department-wide programme later, because the workflow library and governance note are the assets that make scaling cheaper.

Whichever tier you choose, model the full cost including governance and ongoing run costs rather than licence fees alone. The enterprise AI roadmap cost model and the AI implementation cost breakdown for Malaysia both exist to stop the second-year surprise.

Where a pilot usually becomes a custom solution

Around week eight, a pattern tends to appear: the team can draft faster, but the follow-up workflow needs something the tool cannot do — pulling the right account context from the CRM, routing to the right owner, or logging activity automatically. That is the realistic boundary between off-the-shelf software and custom AI solutions, and it is a decision to make after the pilot, not before.

For a sales follow-up workflow specifically, a custom AI chatbot handling enquiries with escalation to a human owner is often the difference between faster drafts and a faster response time measured in the CRM. Keep that conversation out of the 90-day pilot budget so the pilot is not judged against a scope it never had.

What to bring to the approval conversation

Walk in with four things: the two workflow names, the baseline table with numbers filled in, the tier you recommend and why, and the date the decision will be reviewed. That is a defensible ask regardless of how enthusiastic anyone is about the technology.

On the enablement side specifically, AIHQ has trained and engaged over 9,000 professionals and is a registered HRD Corp training provider, with programmes that can be structured to be HRDC claimable subject to client eligibility, grant approval and HRD Corp submission requirements. That experience matters less as a credential than as a signal that someone has run this rollout before and knows where week six stalls.

FAQ

Should the pilot budget sit with marketing, sales or HR? Split it. Tooling and workflow ownership sit with the department; training, baseline design and measurement sit with HR or L&D. A single owner makes the pilot easier to approve but harder to measure honestly.

How many people should be in a 90-day pilot? Aim for 15–30 across both workflows. Fewer gives you noise rather than signal; more makes review load unmanageable and slows the enablement to a crawl.

Is training worth funding if the team already uses ChatGPT? Often yes. Daily personal use and consistent workflow use are different capabilities. Training reframes usage around the two locked workflows and sets a shared quality bar, which is what most pilots actually lack.

What if the pilot shows no measurable improvement? That is a valid result and a cheap one. Document the workflow that failed, the review load it created and the reason, then stop. Pilots that end cleanly protect the credibility of the next one.

Can the training portion be claimed through HRD Corp? AIHQ programmes can be structured to be HRDC claimable, subject to client eligibility, grant approval and HRD Corp submission requirements. Confirm your organisation's eligibility and submission timing with your HR team before committing budget.

Do we need governance work in a 90-day pilot? Yes, but proportionally. A one-page usage note in week three and a short standing policy in week thirteen is usually enough for two low-risk workflows. Heavier governance is appropriate when AI agents or automated actions are introduced.

← Back to all articles