Discovery Call to Production Pipeline: The Ertas Engagement Model

Most enterprise AI vendors make you sit through a demo before they ask what you actually need. You get the slide deck, the product tour, the case study from a Fortune 500 company that looks nothing like your organization. By the time anyone discusses your data, your infrastructure, or your constraints, you have already spent an hour hearing about features you may never use.

We built our engagement model to work the other way around. We start by listening. Here is how the full process works, from first conversation to a production pipeline your team owns and operates.

---

Stage 1: Discovery Call (30 Minutes)

The discovery call is a conversation, not a pitch. No slides. No demo. No sales pressure.

We ask questions:

• What data are you working with? What types, what volume, what formats?
• Where does the data live? Cloud, on-premise, air-gapped?
• What are you trying to build? What model, what use case, what outcome?
• What has been tried before? What worked, what did not?
• What are the constraints? Compliance, budget, timeline, team capacity?

You ask questions too. That is the point. The discovery call exists to determine whether there is a fit — not to convince you there is one.

What happens after the call:

If there is a clear fit, we move to scoping. If there is not, we say so. We have told organizations that their data is not ready for preparation, that their use case does not require a custom pipeline, or that another vendor is a better match. Wasting your time is not a business strategy.

If the situation is ambiguous — you are not sure what you need, or we need to see the data before we can assess — we may suggest a paid discovery session (typically 2-3 days) where an engineer reviews your data environment and delivers a scoping report.

---

Stage 2: Scoping

Scoping translates the discovery conversation into a concrete plan. This is where vague goals become specific deliverables.

A scoping document includes:

• Data sources: Exactly which systems, databases, and file stores will be included
• Pipeline scope: What the pipeline will do (ingest, clean, label, transform, export) and what it will not do
• Deliverables: The specific outputs — a working pipeline, documentation, trained team, quality metrics
• Timeline: Week-by-week plan with milestones
• Team requirements: Who from your organization needs to be available, and when
• Infrastructure requirements: What compute, storage, and network access is needed
• Pricing: Fixed project price or time-and-materials estimate with a cap
• Success criteria: How both sides will know the engagement was successful

Scoping typically takes 3-5 business days after the discovery call. For complex engagements, it may involve a second call or a brief on-site visit.

The scoping document is not a contract. It is a shared understanding of the work. You review it, ask questions, request changes. We iterate until both sides agree on what "done" looks like.

---

Stage 3: Letter of Intent and Engagement Start

Once scoping is agreed, we formalize with a letter of intent (LOI) or a statement of work (SOW). This covers:

• Engagement scope (referencing the scoping document)
• Pricing and payment schedule
• Timeline and milestones
• IP ownership (you own the pipeline and all outputs)
• Confidentiality and data handling terms
• Termination provisions

Payment typically follows a milestone structure: 30% at engagement start, 40% at the build milestone, 30% at handoff. For smaller engagements, it may be simpler — 50% up front, 50% at delivery.

We do not require multi-year commitments. The engagement has a defined scope and a defined end. If you want to extend or add scope later, that is a separate conversation.

---

Stage 4: Forward Deployment

This is where the work happens. An Ertas engineer (or a pair, for larger scopes) embeds with your team.

"Embeds" means different things depending on your environment:

• On-site: The engineer works at your facility, on your network, using your hardware. This is typical for air-gapped or highly sensitive environments.
• Virtual embed: The engineer works remotely but on your infrastructure via secure access (VPN, bastion host, virtual desktop). Daily standups and shared working sessions maintain the collaborative dynamic.
• Hybrid: On-site for critical phases (discovery, domain expert sessions, handoff), remote for build work.

The deployment model is agreed during scoping based on your security requirements, data sensitivity, and practical logistics.

What the Engineer Does

Week 1: Data audit and environment setup. The engineer maps your data landscape, gets access to systems, and confirms (or revises) the scoping assumptions. This is where surprises surface — and they almost always do. The data is messier, more distributed, or more voluminous than expected. That is normal. The engagement plan accounts for it.

Weeks 2-3: Pipeline build. The engineer builds the data pipeline on your infrastructure:

• Ingestion connectors for your source systems
• Cleaning and transformation rules, developed iteratively with your domain experts
• Label schema design and annotation workflow setup
• Quality checks at each pipeline stage
• Export in your required format

Your team participates throughout. Domain experts review output, provide feedback on labels, and flag edge cases. Your engineers observe the build process and ask questions. This is not a black box — when the engineer leaves, your team should understand every component of the pipeline.

Week 4: Validation and handoff. The pipeline runs end-to-end with production data. Quality metrics are measured and reviewed. Documentation is written. Your team is trained on operation and maintenance. The engineer walks through every configuration, every rule, every decision — and explains why, not just what.

---

Stage 5: Training and Knowledge Transfer

Training is not a one-hour walkthrough on the last day. It is woven into the entire engagement.

During the build phase, your engineers work alongside ours. They see how pipeline components are designed, how edge cases are handled, how quality checks are structured. By the time the formal handoff happens, your team has already been learning for weeks.

The formal training covers:

• Pipeline architecture and configuration
• How to modify cleaning rules and label schemas
• How to add new data sources
• How to monitor pipeline health and quality metrics
• Troubleshooting common issues
• How to extend the pipeline for new use cases

We also deliver written documentation: architecture diagrams, configuration references, runbooks, and a troubleshooting guide.

---

Stage 6: Validation and Acceptance

Before the engagement closes, both sides review the success criteria defined during scoping:

• Does the pipeline ingest data from all specified sources?
• Do cleaning rules produce correct output on a validated sample?
• Does the label schema cover all relevant categories?
• Do quality metrics meet the agreed thresholds?
• Can your team operate the pipeline independently?
• Is documentation complete?

If something does not meet criteria, we fix it. The engagement does not close until both sides agree the deliverables are met.

---

Stage 7: Handoff and Post-Engagement Support

At handoff, you own everything:

• The pipeline code and configuration
• All processed data and outputs
• Documentation and training materials
• Any custom scripts or tools built during the engagement

There is no vendor lock-in. The pipeline runs on your infrastructure, uses open formats, and does not phone home. If you never talk to us again, the pipeline keeps working.

Post-engagement support is included for 30 days: email and video call availability for questions, troubleshooting, and minor adjustments. After that, we offer optional support contracts for organizations that want ongoing access.

---

What Happens Next

Some organizations run their pipeline independently from day one. Others come back for follow-on engagements: adding new data sources, building pipelines for additional use cases, or extending the system to new teams.

The engagement model is designed so that each engagement is self-contained. You get a working deliverable at the end, not a half-finished system that requires continued vendor involvement.

---

Starting the Conversation

The first step is always the discovery call. Thirty minutes, no pitch, no obligation. We will ask about your data and listen to what you are trying to accomplish.

Book a discovery call and let us figure out together whether there is a fit.