Playbook

Onboarding an AI agent role looks a lot like onboarding a person, with more precision and better evidence. This playbook takes a Digital FTE from role definition to its first performance review, step by step.

An MCP gateway is where an enterprise turns a protocol into a control. This playbook builds one: identity for agents, delegated user context, per-tool authorization and gates, rate limits, logging, and the registry that makes servers an inventory.

Giving an agent a database connection is the fastest way to get value and the fastest way to get a breach. This playbook builds a Postgres MCP server that exposes safe, permissioned tools instead of raw SQL, with the controls that make it production-grade.

Salesforce is the system agents most often need and the one where a careless integration does the most damage. This playbook builds a thin, permissioned MCP server over the Salesforce API, from tool design to production.

ServiceNow holds the tickets, the CMDB, and the approvals that ITSM agents depend on. This playbook builds a thin, permissioned MCP server over the ServiceNow APIs so agents extend the platform instead of bypassing it.

Browser agents reach the portals and web apps that have no API. This playbook builds one safely: a bounded task spec, an isolated browser with allowlists and scoped credentials, checkpoints and stop conditions, and shadow-mode evaluation before production.

Adoption is happening whether or not there is a plan. This playbook gives engineering leaders a sequence that keeps quality while throughput rises: policy, access, gates, specifications, review, measurement, and expansion codebase by codebase.

Every engineering organization loses hours to questions the codebase could answer: how does this service work, where is that config, who owns this module. This playbook builds an internal developer agent that answers from your repositories and docs, safely.

Switching model providers is a configuration change if the platform was built for it and a rewrite if it was not. This playbook gives the sequence that keeps quality measured at every step: qualify, stage, shift, confirm, retire.

Coding agents can read, explain, and transform legacy code at a pace no team could match, and they can also break it plausibly. This playbook sequences modernization so agents do the volume and verification catches the errors.

Routing is where multi-model strategy becomes real: each request goes to the model that fits its quality, latency, cost, and data constraints. This playbook designs the rules, implements them in the gateway, and keeps them honest with evaluation and cost data.

Most RPA estates have a brittle minority that consumes most of the maintenance budget. This playbook migrates those bots to governed agents, process by process, without a big-bang rewrite and without losing the governance the RPA program built.

A GitHub agent that triages pull requests, assists review against the spec, handles issues, and keeps repositories healthy, under scoped app permissions and the same gates as any coding agent. This playbook builds it from role to production.

A HubSpot agent that keeps the CRM honest, routes leads by rule, turns meetings into records, and drafts outreach for review, without committing, discounting, or emailing on its own. This playbook builds it from role to production.

Engineering teams lose hours to Jira hygiene: triaging vague tickets, chasing missing fields, writing status updates. This playbook builds a Jira agent that triages, enriches, drafts, and reports under scoped permissions, without touching what engineers decide.

Small and mid-sized finance teams run on QuickBooks and spreadsheets. A QuickBooks agent categorizes, prepares, reconciles, and reports, while every posting and payment stays with a person. This playbook builds it with the controls a bookkeeper would insist on.

A ServiceNow agent that fulfills catalog requests, enriches incidents, and answers from your knowledge base, while approvals and change control stay in the platform. This playbook builds it from role definition to production.

A Zendesk agent that resolves the routine majority of tickets with real actions, verifies identity, escalates with full context, and is measured on resolution rather than deflection. This playbook builds it intent by intent.

Chatbots answer; agents resolve. This playbook upgrades a question-answering bot into a governed agent that reads the customer's record, takes actions within limits, and escalates with context, without repeating the mistakes that gave chatbots their reputation.

Basic RAG retrieves once and hopes. Agentic RAG plans, retrieves iteratively, verifies, and knows when to stop. This playbook upgrades an existing pipeline in stages, measured at each step, without rebuilding what already works.

Confluence holds the answers and buries them. This playbook builds a knowledge agent that answers from your spaces under each user's permissions, cites pages, flags stale content, and turns unanswered questions into a documentation backlog.

Gmail, Drive, Calendar, and Sheets hold most of a knowledge worker's day. This playbook builds a Workspace agent that triages mail, drafts replies, finds documents under the user's own permissions, and schedules, with OAuth scopes that keep it honest.

Teams is where employees already ask questions and request things. This playbook builds a Teams agent that answers from your sources, takes bounded actions through governed tools, and runs under the tenant's identity model rather than around it.

NetSuite runs the order-to-cash and procure-to-pay cycles for mid-market companies, and its exceptions run the finance team. This playbook builds a NetSuite agent that resolves exceptions within policy and prepares the close, under the ERP's own role model.