From Legacy CRM to Agentforce in 90 Days

The Problem Nobody Wanted to Admit

When we first spoke to the operations director at a Pune-based industrial components manufacturer, she was candid: "Our CRM is eight years old, our sales team barely uses it, and our service team has never heard of it." The company had roughly 180 employees, a growing export business, and a customer base spanning three continents — all managed through a patchwork of spreadsheets, WhatsApp threads, and an on-premise CRM that hadn't been meaningfully updated since 2016.

This is more common than most enterprises want to admit. The CRM gets implemented, adoption stalls, and workarounds proliferate until the workarounds become the actual system. By the time leadership decides to fix it, the gap between what the CRM holds and what's actually happening is enormous.

The trigger for this client was simpler: they'd lost a major account renewal because a relationship change at the customer's side wasn't captured anywhere. The account manager had left six months prior. Nobody knew.

Why 90 Days and Not 18 Months

Enterprise transformation projects earn their bad reputation honestly. Long timelines, ballooning scope, and the dreaded "big bang" go-live that never quite works. We proposed something different: a phased, outcome-first approach anchored on 90-day delivery cycles with working software shipped at each milestone.

This distinction matters. A technology-led project optimises for features. An outcome-led project optimises for adoption and business impact. Every decision we made — what to configure, what to defer, what to integrate — was filtered through that lens.

Phase 1: Foundation (Weeks 1–4)

We started by conducting account and contact data archaeology. Six years of historical data across three spreadsheet formats, two old CRM exports, and an Outlook shared mailbox. We mapped, deduped, and loaded clean records into a new Salesforce Sales Cloud org — not everything, just the active accounts, open opportunities, and the last 24 months of won deals. Clean data to start, then migrate history incrementally.

Simultaneously, we configured the core Sales Cloud objects to match how the team actually works — not how a generic CRM implementation guide says they should work. Custom stages, the right required fields (and only those), and a mobile layout that makes sense for reps who spend most of their time on-site with clients.

Phase 2: Process Automation (Weeks 5–10)

With a working CRM in place, we layered in automation. Flow Builder replaced three manual email sequences. An assignment rule handled territory routing automatically. A renewal forecast dashboard surfaced upcoming renewals 90 days out — the exact failure point that triggered this project.

We also connected Salesforce to the company's ERP via MuleSoft, creating a live sync of order history into the account timeline. For the first time, account managers could see every purchase, shipment, and support case in a single view.

Phase 3: Agentforce (Weeks 11–14)

Here's where things got genuinely exciting. With clean data, structured processes, and an integrated ERP, the foundation for AI was solid. We deployed an Agentforce agent scoped to two specific jobs: answering common account status queries from the internal team, and drafting renewal outreach emails based on account history and the rep's input.

We deliberately kept the agent narrow. Not because broader wasn't possible — it was — but because a focused, reliable AI tool builds trust faster than an ambitious, unpredictable one. The renewal email drafts alone saved each rep approximately 2.5 hours per week.

What Made the Difference

Three things stand out from this engagement that we'd carry into any similar project:

• Data first, features second. We spent more time on data quality in the first two weeks than most teams spend in an entire implementation. It paid dividends at every subsequent phase.
• No configuration for configuration's sake. Every field, every automation, every permission set was justified by a real user need. The out-of-box complexity was our enemy, not our ally.
• The agent came last — and that was intentional. Agentforce is powerful, but it's only as good as the data and processes underneath it. By the time we introduced AI, the team already trusted the system it was drawing from.

What 90 Days Actually Looks Like

Fourteen weeks of focused work, weekly checkpoints, and no big surprises. The team went from spreadsheet-dependent to Salesforce-native in the time it takes most firms to finish discovery. The Agentforce agent is now handling roughly 40 account queries per day with a 94% satisfaction rating from internal users.

The operations director's summary at the retrospective: "I thought this would take two years and feel like a punishment. It took three months and felt like we finally have a system that works for us."

That's what outcome-led delivery looks like in practice.