24 posts tagged “IBM Bob”

How a Maximo development team builds an AI Layer on top of its customisation repo so any coding agent (Claude Code, IBM Bob, Codex) writes autoscripts, Java MBO extensions, DBC scripts, MAF apps and integrations the house way: lean global rules, per-artifact rules, a command library (/prime, /plan-autoscript, /validate), hooks, validation against a dev Manage instance, a separate reviewer, autonomy levels per artifact, parallel work, and system evolution in a 15-year-old customised codebase.

How to make IBM Bob write Maximo code your way: the rule load order across .bob/rules, rules-{slug}, and AGENTS.md, a full Maximo AGENTS.md, two custom modes with fileRegex and allowedSubagents, a Maximo skill built on the anatomy of IBM's maximo-script-optimizer, slash commands, plugins with mcp.json, RequiredExtensions, and IBM's own ibm-mas/cli setup.

How to put deterministic guardrails around IBM Bob on a Maximo estate: the lifecycle hooks schema and its seven documented events, the new 2.3.0 permission events, four illustrative Maximo guardrail hooks, the EnforcedHooks group policy, OpenTelemetry policies, a least-privilege .bob/mcp.json for the MAS 9.2 Maximo MCP server and agent-neutral servers, workspace trust, .bobignore, and a layered-defense table that says what each layer stops and what it does not.

How to run IBM Bob Shell headless in CI for a Maximo customization repo: install, bob run flags and JSON schemas, cost and turn caps, IBM's own galaxium-travels review workflow, an illustrative GitHub Actions reviewer for automation scripts and app XML, MXAPIAUTOSCRIPT export, policy and MCP OAuth limits, and IBM's productized pull request reviews.

A Maximo reading of everything IBM published for Bob between the self-hosted release and 10 November 2026: the four premium packages (DevSecOps, Java Modernization, Z, i), how the model router picks a tier, IDE 2.3.0, what TechXchange did and did not say, the Maximo assets that actually changed, a dated scorecard, and a refreshed open-questions checklist.

How to pilot IBM Bob on a Maximo estate without trusting a vendor number: IBM-reported customer metrics versus independent views, the competitive field, eight points of guidance, a pilot metric design (defect escapes, review hours, Bobcoins per task), and the open questions as a checklist.

What IBM Bob can actually do against Maximo today: the MAS 9.2 native MCP server, IBM's two Maximo Bob skills, MAF Local Dev Mode, a community evaluation, a task-fit table, and where TMG's Max_Interfaces, Max_mcp, and Max_autoscripts fit.

What IBM Bob's group policies, Bobalytics, CADF activity log, SIEM feed, SSO, plans, and Bobcoins actually enforce, where each control stops, and how to wire them into a Maximo change advisory board so AI-authored changes stay auditable.

SaaS Bob auto-routes models and you cannot pin one. Self-hosted Bob runs a single core model on OpenShift. Here is what IBM states about data handling, indemnity, regions, the PromptArmor incident, sizing, and GA limits, with every unverified claim flagged.

IBM Bob started as an internal pilot and is now a SaaS and self-hosted agentic platform. Here is the dated timeline, the V2 three-tier design, the clients, the Agent/Plan/Ask modes, and the extension points a Maximo team will actually touch.

A nine-part, vendor-number-labeled field guide to IBM Bob for Maximo shops: what Bob is, how its models and security work, how to govern and price it, how it meets Maximo through MCP and skills, and how to pilot it without trusting a single marketing figure, plus house rules, guardrails, CI and what IBM shipped through November 2026.

AI coding agents have arrived in IBM Maximo land: a native MCP server in MAS 9.2, Maximo skills for IBM Bob, and open tooling from the community. Here's what changes for Maximo teams, what doesn't, and where the guardrails go.

An AI coding agent walks into your codebase knowing nothing about your naming conventions, banned libraries or deployment rules. Here's how AGENTS.md, rule files, skills and MCP turn house rules into something the agent actually follows.

AI coding vendors love big productivity numbers. Here's how to read them like a skeptical executive, what the IBM Bob figures actually measure, and how to run a pilot that produces a number you can trust.

When an auditor asks who wrote that line of code, "the AI did" isn't an answer. How enterprise teams keep receipts for AI-assisted development: activity logs, usage analytics, approvals, rollbacks, and agents that act through APIs instead of databases.

Autocomplete, legacy refactoring, and architecture planning don't need the same AI model. Here's how enterprise coding tools decide which model does which job, and why letting the tool choose for you has a catch.

For banks, utilities, defence and government, 'just use the cloud' isn't an answer. Here's what self-hosted and air-gapped AI coding really means in 2026 — and the capability trade-off nobody puts on the slide.

Forget greenfield demos. The most valuable thing AI coding agents do in banks, utilities and government is read, document and modernize the COBOL, RPG and old Java nobody wants to touch — with real caveats.

Approval prompts feel like friction — until a README tricks an AI agent into running malware. What the January 2026 IBM Bob beta incident teaches every enterprise about where humans belong in AI coding.

Cursor, Claude Code, Kiro and IBM Bob all ship some version of Ask, Plan and Agent modes. Here's why the 'think first, type second' habit is the single cheapest guardrail in enterprise vibe coding — and when to skip it.

AI coding went from finishing your line, to answering your questions, to doing the work across your whole system. Here is what changed in each era, why agents need a way into enterprise systems, and how the Model Context Protocol became that way in.

When an AI agent writes the code, your prompt is the requirement. A vague one gets you code that works and is wrong. Here is how enterprise teams are turning prompts into real specifications, with examples from Kiro, IBM Bob, Claude Code, Cursor and Copilot.

Vibe coding started as a throwaway tweet about letting the AI drive. Eighteen months later it is inside banks, utilities and government agencies, wearing a navy blazer. Here is what changed, what didn't, and why even IBM won't call it vibe coding.

A 13-part, plain-English series on what happens when vibe coding leaves the weekend side project and walks into banks, utilities, insurers, government agencies and manufacturers. Specs, agents, approvals, audit trails, legacy code, firewalls, and how to read the productivity numbers.