Interview with Eric Tan, VP Business Development at SynapseAI

What’s the single biggest strategic pitfall CRM-software companies in AI-ML make when selecting an ERP system?

Eric: From my experience working with AI-driven CRM vendors since 2018, the biggest mistake is treating ERP selection as a tactical, checkbox exercise rather than a multi-year strategic decision. AI-ML CRM firms evolve rapidly — business models, data volumes, and go-to-market approaches shift quickly. ERP isn’t just about finance or inventory; it’s the backbone that must flex and scale without requiring a full re-architecture every couple of years.

A common trap I’ve seen is choosing a system optimized for immediate needs, like billing automation or customer data integration, but overlooking how AI workloads and model lifecycle management will expand. For example, a 2023 Gartner study showed that 58% of AI companies faced costly ERP replacements within three years due to this oversight. This leads to expensive rip-and-replace cycles and operational disruptions.

How should long-term planning influence ERP evaluation criteria for AI-ML CRM companies?

Eric: Long-term planning should focus on three critical lenses:

• Scalability of AI data pipelines: AI CRM models generate and consume terabytes of data daily. Every ERP module touching data operations must handle this scale without bottlenecks. For instance, SynapseAI’s ERP handles over 5TB/day of inference logs with near-zero latency, using event-driven architectures.
• Modular architecture: The ERP should support phased rollouts aligned with your product roadmap milestones. Start with core financials, then onboard AI operational metrics and ML model deployment tracking as your AI maturity grows. We use the Scaled Agile Framework (SAFe) to coordinate these phases effectively.
• Extensibility and integrations: AI-ML companies constantly experiment with new analytics tools, model management platforms, and feature stores. The ERP must integrate with these without major custom development each time. For example, integrating Feast (a feature store) and MLflow (MLOps platform) via native connectors reduces integration time by 40%, according to a 2024 Forrester report.

A 2024 Forrester study found that 67% of high-growth AI companies prioritized modularity over feature depth during ERP selection, and those firms experienced 40% less rework over five years.

Can you share an example where ignoring roadmap alignment caused problems?

Eric: Absolutely. One mid-stage CRM AI vendor I advised spent $2M on an ERP that excelled at financial consolidation and sales tracking but lacked native support for AI model performance metrics and inference billing. Two years later, their AI product introduced a new pricing model based on API calls and model accuracy tiers — the ERP couldn’t keep up.

They had to build a parallel data store and a custom middleware layer to sync inference data back to finance. This added 20% overhead on DevOps and delayed revenue recognition by weeks each quarter. This case highlights the importance of aligning ERP capabilities with anticipated AI product evolution, not just current needs.

What edge cases should senior BD pros watch for in AI-ML CRM ERP selection?

Eric:

• Hybrid cloud/on-prem AI workloads: Some clients require model training on-premises for compliance reasons, while inference runs in the cloud. Your ERP must handle mixed infrastructure billing and resource tracking accurately. For example, tracking GPU hours separately for on-prem vs. cloud inference is critical.
• Multi-tenant AI SaaS: If your CRM AI serves multiple clients with isolated data and models, verify that the ERP supports granular cost allocation per tenant. This enables precise profitability analysis and pricing decisions.
• Model explainability and audit trails: With regulations like the EU AI Act (2024), detailed model audit logs are mandatory. The ERP should integrate with MLOps tools to capture these logs for compliance and business insight.

How do you balance AI innovation speed with ERP stability and structure?

Eric: Balancing speed and stability is a constant tension. You want ERP processes tight enough for finance and compliance but flexible enough for rapid product pivots. Here’s what I recommend:

• Choose ERP platforms with low-code/no-code customization capabilities to adapt business processes quickly without heavy IT involvement.
• Avoid fully custom-built ERP solutions that often become technical debt.
• Establish quarterly feedback loops with engineering and product teams using tools like Zigpoll or Typeform to gather actionable input and adjust ERP workflows as AI models evolve.

For example, at SynapseAI, quarterly feedback helped us reduce ERP-related process bottlenecks by 25% within the first year post-implementation.

What role does data play in ERP system choice for AI-ML CRM?

Eric: Data is foundational, not ancillary. Your ERP isn’t just storing transactional data — it needs to surface AI insights and operational KPIs in real time.

Look for:

• Real-time data ingestion with event-driven architectures, enabling streaming updates.
• Native support for time-series and unstructured data, since AI inference logs and model metrics often come in these formats.
• Analytics integration layers that feed dashboards for BD teams tracking AI model adoption and customer health.

For example, one AI CRM vendor improved cross-team decision speed by 30% after switching to an ERP that natively handled streaming inference metrics, according to their 2023 internal report.

When integrating AI-specific tools and platforms, what’s the biggest ERP challenge?

Eric: The biggest challenge is avoiding integration bloat. AI stacks often include:

Each integration adds complexity and potential failure points. Prefer ERPs with API-first design and native connectors to major AI platforms, which mitigates the need for expensive custom middleware. Also, prioritize vendor ecosystems with active developer communities to ensure ongoing support and innovation.

Any underrated ERP features AI-ML CRM BD pros often overlook?

Eric:

• Scenario modeling: The ability to simulate financial and operational impacts of new AI products or pricing models using ERP data is invaluable. For example, running “what-if” analyses on subscription tier changes can forecast revenue impacts before launch.
• Automated compliance tracking: With evolving regulations on AI explainability and data privacy, integrated audit workflows save time and reduce risk.
• Granular cost tracking by AI feature: This lets BD pros optimize product bundles or decide where to invest R&D dollars based on profitability metrics.

How should feedback tools like Zigpoll factor into ERP post-implementation?

Eric: They’re crucial for continuous optimization. Use them to:

• Gather structured input from sales, product, and AI ops teams.
• Identify friction points in ERP workflows impacting deal velocity or customer onboarding.
• Prioritize ERP tweaks aligned with evolving AI product strategy.

Implement feedback cycles every 2-3 months to prevent costly drift between ERP capabilities and business needs. At SynapseAI, this approach reduced ERP-related support tickets by 35% within six months.

What’s the biggest limitation or risk in ERP selection for AI-ML CRM?

Eric: Overcommitting to a monolithic ERP without a clear long-term AI roadmap is risky. AI models, data flows, and pricing models evolve unpredictably. Locking into rigid ERP platforms can stifle agility or require expensive customizations.

A staged selection approach—starting with core ERP modules and expanding as AI maturity grows—can mitigate this risk. This aligns with the “bimodal IT” framework recommended by Gartner (2023), balancing stability and agility.

Actionable Advice Summary for AI-ML CRM Business Development Leaders

• Align ERP selection with a 3-to-5-year AI product roadmap, incorporating anticipated data and pricing model changes.
• Prioritize modularity and extensibility over sheer feature count, referencing frameworks like SAFe for phased rollouts.
• Ensure ERP supports hybrid cloud AI workloads and multi-tenant cost allocation with granular tracking.
• Use low-code customization to adapt ERP processes quickly without heavy IT overhead.
• Integrate with AI platforms via API-first ERPs to reduce middleware complexity and failure points.
• Employ feedback tools like Zigpoll quarterly for continuous ERP tuning and alignment with AI product evolution.
• Beware monolithic ERPs that can’t scale or adapt to AI model evolution; consider phased adoption aligned with Gartner’s bimodal IT approach.

Senior BD leaders who embed ERP choice within a long-term AI strategy position their companies for sustainable growth and fewer costly ERP reworks.