Start with the work that eats your sales team's week but never closes a deal: logging calls and updating records, chasing quotes that went quiet, answering the same first questions on WhatsApp, and sorting serious leads from tyre-kickers. These tasks are repetitive, follow clear rules and are easy to check, so they're the safest place to let AI take over. Keep pricing negotiations, complaints and relationship calls with your people until the routine work is running reliably.
That routine work is a bigger share of the job than most owners assume. Salesforce's 2023 State of Sales report, which surveyed more than 7,700 sales professionals across 38 countries, found reps spend only about 28% of their week actually selling. Most of the rest goes on admin, data entry and managing deals.
Why the order you automate in matters
Most businesses that waste money on AI automation pick something visible, not something costly. A WhatsApp auto-reply sounds impressive until a bot answers so flatly the lead goes cold.
The right question isn't "what can we automate?" It's "what is costing us deals right now because we're slow or inconsistent?" The four tasks below meet that test.
1. Enquiry triage
Enquiries arrive from WhatsApp, email, website forms and Instagram DMs at the same time. When the person responsible is busy, enquiries wait, and speed matters more than most owners think. A study reported in Harvard Business Review (Oldroyd, McElheran and Elkington, 2011) found firms that responded to an enquiry within an hour were nearly seven times as likely to qualify the lead as firms that waited even one hour longer.
An AI agent handles this the moment an enquiry arrives: it reads the message, checks for key signals (budget, timeline, service named), and either routes it with a summary or sends a qualifying reply to gather missing information. For most SMEs, a rules-based agent (Tier 1, starting from RM1,000) handles the majority of it.
2. Quote follow-up
You sent the quote. Three days pass. The prospect buys from whoever followed up first. Quote follow-up is almost always done inconsistently: different staff follow different schedules, and some don't follow up at all.
An AI agent watches your quote pipeline, triggers a follow-up at a set interval, logs the response, and escalates to a human when the prospect replies with an objection or buying signal. If you're using Xero or QuickBooks, the agent can pull the quote date automatically; HubSpot or a spreadsheet works too with different connectors.
3. Lead qualification
A qualification agent runs your standard questions automatically (budget, timeline, decision-maker), scores the lead against your criteria, and passes only qualified leads to your sales team. In practice the rule can be as simple as: budget stated, timeline under three months, and talking to the person who signs off means the lead goes straight to sales; anything missing gets a follow-up question first.
This is where the data from the real-cost-of-manual-work article becomes concrete: hours spent on bad leads are hours not spent closing good ones.
4. CRM data entry
After every call or site visit, someone is supposed to log the outcome. In practice it gets done at the end of the week, from memory, with gaps. A CRM that's only partly accurate is worse than useless for forecasting.
An AI agent can transcribe a call summary, extract key fields (contact name, company, next action, deal stage), and push them to HubSpot or your CRM automatically, without your sales team touching a keyboard after every call.
Which task should you automate first?
| Task | Automate first if… | Do not automate yet if… | Typical human checkpoint |
|---|---|---|---|
| Enquiry triage | More inbound messages than your team can answer within two hours | Volume is low enough one person handles it comfortably | Escalation when signals indicate a high-value lead |
| Quote follow-up | Quotes regularly go unanswered for more than three days | Your team follows up reliably and fewer than five quotes are open at a time | Agent escalates on objection or buying signal |
| Lead qualification | Sales team spends significant time on discovery calls that don't convert | You lack clear, consistent qualification criteria | Human reviews all borderline scores |
| CRM data entry | Pipeline data is incomplete or updated only weekly | CRM not yet configured with standard fields and deal stages | Sales manager spot-checks entries weekly |
What not to automate first
Negotiation and objection handling require judgment that changes with every client, so keep a human there. Proposal writing is only viable after you've built a clean data source (service catalogue, pricing logic, standard scope items); without it, the agent produces drafts your team rewrites entirely. Pricing approval is always a human decision; an agent can surface the request and context, but should not approve a discount.
Compliance and data
If your AI agent handles prospect data, that's a personal data question under Malaysia's Personal Data Protection Act (PDPA). Pexalo designs agents with PDPA- and GDPR-aligned controls, including action logs, role-based access and human approval checkpoints.
On AI specifically, Malaysia's Ministry of Science, Technology and Innovation (MOSTI) published the National Guidelines on AI Governance and Ethics in September 2024. They're voluntary for now, but they set out the principles (fairness, transparency, accountability, privacy) that future rules are likely to build on. Keep an eye on updates from the Personal Data Protection Department (JPDP) as guidance develops.
Your client data stays yours. We use business API terms where providers don't train their models on your data, and we can confirm which providers a build uses before you sign.
Where to start
Every engagement starts with a free workflow audit that tells you which of the four tasks costs you the most, what a realistic build looks like, and whether Tier 1 or Tier 2 is the right fit. See how a workflow audit actually works before you book one.
FAQ
Where do I start if I'm not sure which task hurts most? Book the free audit. It maps your workflow and ranks the gaps before you commit to anything.
Can an agent handle price objections? No. An agent escalates when a prospect raises an objection; a human handles the commercial conversation.
What if my CRM isn't set up properly yet? Configure standard fields and deal stages first. An agent logging data into an inconsistent schema makes forecasting harder, not easier.
Is my prospect data safe? Pexalo builds with PDPA- and GDPR-aligned controls, logs every action the agent takes, and uses business API terms where providers don't train their models on your data.
Start with enquiry triage if leads are going unanswered. Start with quote follow-up if your close rate is lower than it should be for the volume you're sending. Start with lead qualification if your team spends too much time on conversations that go nowhere. Any of those will pay back faster than automating whatever looks most impressive.
Book a free workflow audit and find out which one will save you the most.