Skip to content

Implementation lifecycle

superglue follows the steps an experienced consultant already uses, from the implementation plan to the handover. The agent does the reading, drafting, data and configuration work. You make the decisions and grant every permission to write to a client system.

An implementation has 11 steps in three phases. Each step has its own page.

Phase Step What the agent produces How to run it
Pre-implementation 1. Implementation plan Project with tasks, owners and dates Project setup
2. Discovery Requirements, decision log, open questions Discovery
3. Data assessment Inventory and readiness report Data assessment
4. Solution design Fit-gap matrix and design document Solution design
Implementation 5. Configuration Applied settings and a configuration workbook Configuration
6. Data mapping Field mapping and transformed files Data mapping
7. Data migration Loaded records and reconciliation Migration and cutover
8. Testing and UAT Test scripts, UAT tracker, defect tasks Testing and UAT
9. Cutover Cutover runbook, production and delta loads Migration and cutover
Post go-live 10. Hypercare Scheduled syncs and failure alerts Post go-live
11. Handover Closing documents and the audit trail Handover

Connect the client’s systems when a step first needs them. Private and on-premise systems connect through the secure gateway. See Connections.

The setup guides say what to ask the client for, for each system: NetSuite, Sage Intacct, Acumatica, Business Central, QuickBooks Online, QuickBooks Desktop, Xero, Sage 100 and Sage 300.

The steps are the same for every system, but the decisions differ. The website has a page per step for the most common targets:

  • Approvals: The agent reads source and target systems without asking, for reads it recognizes such as GET requests and single SELECT queries. By default, before it writes, updates or deletes data in a client system, it lists the permissions it needs once in plain language, such as “Create 1,240 vendors in NetSuite”, and waits for your approval. A session switched to Act without asking grants its own requests.
  • Runs: Loads and syncs run as saved tools with no AI at execution time, so the sandbox load and the production load apply the same transformation. Every run is recorded, and each step’s result and the output are kept when Store run results is on. See Monitoring runs.
  • Playbooks: Ask the agent to copy a built-in playbook and adapt it to your firm’s method, so every consultant runs implementations the same way. See Playbooks.

Start with Project context. If you only want to try the agent on one system first, a Data assessment is a safe, read-only starting point.