Home / Wiki / AI agents

AI agents

99 items across the archive · 31 entries · 37 assets · 18 papers · 7 courses · 6 writings

Systems where a model plans, calls tools and acts — how they are built, where they fail, and what it takes to run them for real users.

What the archive holds on this

Everything on this page mentions ai agents directly. It is drawn from 99 items across every part of Khurram Badar's archive — working entries, the reusable frameworks and checklists he calls moats, full papers and course material, and published writing — and each links to its source page.

Working entries

2026-08-13 · government policy

Dubai agentic AI mandate strategy

Khurram mapped Dubai's May 2026 Sheikh Hamdan agentic AI mandate, identified TDRA cold-call ban (AED 5K fines), and chose referral partnerships over walk-ins. The value: multilingual ASR capability (Urdu/Hindi/Arabic/English) defensible for legacy-stack integration that Western vendors miss.

For a board director
This gives you a compliant market-entry route into Dubai's agentic AI mandate that avoids AED 5K fines from the TDRA cold-call ban.
For a CEO
You get a referral-based sales strategy for Dubai's AI mandate that sidesteps regulatory penalties while leaning on multilingual ASR as a differentiator against Western vendors.
For an HR / people lead
You should note the referral-partnership approach shapes how sales staff engage Dubai prospects, keeping the team clear of the TDRA cold-call ban and its fines.
For a finance lead
You can weigh referral partnerships against walk-in costs, factoring in the AED 5K fine risk the cold-call ban creates if that channel were used instead.
2026-07-11 · legal compliance

Mizan Legal AI Assistant Platform

Khurram built a presentation deck positioning Mizan as 'GMALC's own AI Assistant Lawyer' with stages including live reception chatbot, agentic intelligence for daily laws/judgments, internal orchestration, fee engine, and staff dashboards. The value: Owned, not rented—an asset that compounds with organizational knowledge versus subscription model, differentiating against cloud-only competitors.

For a board director
This gives you an owned AI asset that compounds internal knowledge over time instead of a recurring vendor dependency, a structural differentiator versus cloud-only competitors.
For a CEO
You can position this platform as a proprietary lawyer-assistant capability spanning reception, daily legal intelligence, orchestration and fee management, distinct from off-the-shelf subscription tools.
For an HR / people lead
Staff dashboards built into this platform give you visibility into how the assistant supports daily work across reception and internal case handling.
For a finance lead
Owning this platform rather than renting it changes your cost structure from ongoing subscription fees to a built asset that accumulates value with use.
2026-07-08 · fintech payments

Metals Trading Desk pre-seed fundraising

researched Money Expo Abu Dhabi 2026 (45 speakers in trading/crypto/fintech) and revealed raising $250K pre-seed for dubaiaihouse.com and metalstradingdesk.com; decided tactical approach: pitch Metals Trading Desk (not portfolio), 30-second opener ('multi-agent AI gold/silver signals across COMEX/London/Shanghai'), target family offices, WhatsApp follow-up with contextual reference, skip printed cards. The value: demonstrates founder discipline—stays focused on single platform (Metals Trading Desk) despite 21 live

For a board director
You should note the decision to pitch one focused product—Metals Trading Desk—rather than the full 21-product portfolio when raising from family offices.
For a CEO
You get a tested pitch structure: a 30-second opener on multi-agent AI signals, targeting family offices, with WhatsApp follow-up instead of printed cards.
For an HR / people lead
You see a founder pattern of risk-aversion-then-commit under pressure, useful context if you're assessing decision-making style for future hires or leadership fit.
For a finance lead
You should track the $250K pre-seed target and the cost-saving call to skip printed materials in favor of WhatsApp follow-up during fundraising.
2026-07-08 · security privacy

CyberKnight AI-native security operations

analyzed CyberKnight (MENA Zero Trust VAD, 60+ vendors) for three AI value-add strategies—portfolio intelligence RAG, gap-assessment tool, agentic SOC platform—then decided strongest bet is AI-native security operations for GCC enterprises in Arabic, not vendor cloning. The value: rejects incremental 'layer AI on incumbents' thinking, believes future belongs to building native products now, not wrapping legacy—direct conviction: 'New products are coming, what can we do now?'—positioning Arabic/GCC/compliance contex

For a board director
This entry flags a strategic bet you should weigh: committing to a new Arabic-native security product for GCC enterprises instead of a safer incremental play on existing vendors.
For a CEO
Your call here is whether to build an AI-native SOC platform for Arabic-speaking GCC customers now, ahead of global competitors localizing, rather than layering AI onto 60+ existing vendor products.
For an HR / people lead
If this direction moves forward, you will need to plan for hiring people with Arabic-language and GCC compliance expertise to build and support a genuinely new product.
For a finance lead
This choice trades the lower-risk path of enhancing existing vendor relationships for the higher-cost, higher-upside route of funding a new product build targeting an underserved Arabic/GCC compliance niche.
2026-07-06 · tooling devex

Ai learning

The way ai connect to your tools your data and systems is MCP Ai integration Agents and applications that use llm Llm integrate unstructured data and ask questions Which fiction to call Which parameters to set Cloud security Models autonomously discover and use tools Api isnthe traditional way software communicates Ai wirks through cobtext Standard way to discover what a tool can do what input it expect and what output it returns all through cobtext Create maxhine readable map of your api Auth Rate limits Creat a c

For a board director
Adopting a standard protocol like MCP for connecting AI to your systems affects vendor risk and long-term interoperability decisions your board should track.
For a CEO
Understanding MCP as the semantic layer between models and your tools helps you decide how fast to build agentic capability without locking into bespoke integrations.
For an HR / people lead
As AI agents get plugged into more internal tools and data sources, you'll need new guardrails and permission layers around who and what can access employee and company information.
For a finance lead
Token infrastructure and rate limits directly affect your AI operating costs, and designing loops that avoid repeated inference calls can meaningfully reduce spend.
2026-06-25 · learning masterclass

Dubai AI initiatives 16-page handbook

rejected surface-level training and conducted exhaustive 12-month sweep of Crown Prince LinkedIn, PM office announcements, and Gulf News to build 16-page handbook covering AI basics, Dubai's AED 100bn annual target, sector-by-sector initiatives (RTA 81 projects, DEWA green data centre, K-12 mandatory AI), and 22 Chief AI Officers + agentic AI program covering 295,000 companies. The value: provides comprehensive curriculum for training government staff and institutions on Dubai's AI vision with verified sources, not

For a board director
This gives your board a verified, source-checked view of Dubai's city-wide AI targets and initiatives to benchmark strategic alignment against.
For a CEO
You get a sector-by-sector map of Dubai's AI initiatives, from RTA to DEWA, to decide where your operations need to align or compete.
For an HR / people lead
This handbook gives you a verified curriculum for training staff on Dubai's AI vision, including the mandatory K-12 and Chief AI Officer structures shaping talent expectations.
For a finance lead
You can use the AED 100bn annual target and sector initiative details to assess where government-driven AI spending may create budget or partnership implications.
2026-05-25 · tooling devex

Kiro.dev vs Claude Code IDE

Evaluated kiro.dev (AWS agentic IDE with spec-first discipline) against Claude Code's conversational workflow and terminal access. The value: Confirmed preference for Claude Code's flexibility for multi-threaded project management (Tarbiyah, Subject Explorer, rapid parallel builds) over rigid specification process.

For a board director
You gain assurance that your teams are choosing development tools based on tested fit for parallel project delivery, not default vendor adoption.
For a CEO
This confirms your teams can keep running multiple builds like Tarbiyah and Subject Explorer at once, using the more flexible of two evaluated IDE options.
For an HR / people lead
This entry is about tool selection for developers and offers little direct relevance to your people or culture priorities.
For a finance lead
Sticking with the more flexible tool over a rigid alternative avoids the cost of forcing teams into a slower, spec-heavy process across parallel projects.
2026-05-11 · tooling devex

Video understanding MCP integration

Khurram clarified his need for video input that feeds into Claude Code for agentic actions, receiving setup for mcp-gemini-video-understanding MCP server to bridge Gemini's video comprehension into Claude Code workflows. The value: Enables video comprehension as a primary input to agentic workflows, expanding Claude Code's sensory surface beyond text for richer operational context.

For a board director
This expands the range of inputs your operating team can act on, widening the technical capability base underpinning agentic workflows.
For a CEO
This gives your team a working setup to feed video directly into Claude Code, adding a new input type your agentic workflows can act on.
For an HR / people lead
This tooling change lets technical staff extend agentic workflows to video input, a skill worth noting for role and training planning.
For a finance lead
There's no cost or spend figure here, but this integration adds a new input capability to existing agentic tooling at no stated additional expense.
2026-05-11 · government policy

Bilal Bin Saqib regulatory AI

Khurram analyzed Bilal Bin Saqib (Pakistan's PVARA chairman) by studying his recent posts and discovered a pivot from crypto regulation into AI builder positioning through technical multi-agent system writeups, mapping a warm-introduction connection strategy. The value: Identifies a bridge figure between frontier AI and Pakistani state, recognizing domain arbitrage opportunity between Western capital, AI innovation, and South Asian regulatory authority.

For a board director
This flags a regulatory bridge figure whose pivot into AI positioning could shape how Pakistani authorities engage frontier AI ventures you oversee.
For a CEO
Understanding this connection strategy gives you a warm-introduction path to a regulator-turned-AI-builder linking Western capital and South Asian regulatory authority.
For an HR / people lead
This has limited direct relevance to your work, though it signals how technical positioning can shift a regulator's professional identity and network.
For a finance lead
This points to a domain arbitrage opportunity between Western capital and South Asian regulatory authority that could inform where you direct expansion capital.
2026-05-05 · tooling devex

AI sidecar specialist positioning

Khurram debated whether Dubai's 50% agentic AI mandate favors his Node.js/Vercel/Supabase stack over Laravel, concluding that agentic workflows run natively on his stack while Laravel shops need a polyglot sidecar approach. The value: Identifies a market segment where he becomes indispensable within enterprise bids as the 'AI sidecar specialist,' competing not on general backend development but on regulatory-grade agentic systems.

For a board director
This positioning gives you a clearer view of how a regulatory mandate is being turned into a differentiated bid strategy against Laravel-based competitors.
For a CEO
You gain a niche 'AI sidecar specialist' pitch to win enterprise bids where the Node.js/Vercel/Supabase stack meets agentic AI requirements natively, unlike polyglot Laravel setups.
For an HR / people lead
You may need to identify or hire talent who can credibly deliver regulatory-grade agentic systems if this positioning becomes central to bids.
For a finance lead
You should note this as a potential revenue angle for enterprise bids that avoids the added cost of building polyglot sidecar architectures competitors need.
2026-05-03 · government policy

FTA Trust Architecture agentic

Khurram created a 43-page whitepaper called 'The FTA Trust Architecture' positioned around Sheikh Maktoum's directive to convert 50% of government operations to agentic AI within two years, spanning four core components: Agent Access Charter, Document Sovereignty Layer, Living Knowledge Graph, and Autonomy Ladder. The value: Positions himself as an 'operating system and constitution for agentic government' rather than selling tools, differentiating through demonstrated regulatory/compliance frameworks understanding

For a board director
Weigh this against your governance risk exposure: a defined framework for agent access, document control, and staged autonomy gives you a structure to evaluate before any agentic mandate is adopted.
For a CEO
If your organization faces pressure to convert operations to agentic AI, this framework offers you a staged path—Agent Access Charter, sovereignty controls, and an Autonomy Ladder—rather than a single tool purchase.
For an HR / people lead
The Autonomy Ladder component matters to you because it defines how roles and decision rights shift in stages as agentic AI takes on operational work, affecting workforce planning.
For a finance lead
Before committing budget to agentic AI adoption, this whitepaper's compliance and access-control framework helps you assess governance costs and risk exposure ahead of large-scale rollout.
2026-04-29 · government policy

Harness-first architecture and FDA regulatory positioning

analyzed Claude Code's 98/2 infrastructure-to-AI split as regulatory language asset—mapped harness architecture (Surface → Core → Safety) to FDA sovereignty requirements (data residency, audit trail, deterministic permissions, logged execution). The value: demonstrates how to reframe AI tooling from 'risky black box' to 'controlled agentic system' when pitching regulated buyers—bureaucratic confidence comes from architecture transparency, not model capability hype.

For a board director
This gives your board a governance narrative that reframes AI adoption as controlled architecture rather than unproven capability, easing regulated-market approval.
For a CEO
You can pitch regulated buyers by presenting your system's data residency, audit trail, and permissions structure instead of relying on model performance claims.
For an HR / people lead
This offers a template for explaining AI tooling to compliance-sensitive stakeholders in terms of controlled processes rather than opaque automation risk.
For a finance lead
You gain a positioning approach that could shorten regulated sales cycles by addressing sovereignty and audit requirements upfront rather than through capability marketing.

Frameworks, checklists and positions (37)

Papers and courses

Written work

Vocabulary

agentic
AI agent systems and autonomous workflows; foundation of modern Claude-based applications.36 documents
autonomously
Self-directed action without external control, capability of AI agents in task completion.32 documents
converting
Transforming business model or focus to new direction like AI agent solutions.31 documents
createagent.ai
Live platform for creating and deploying AI agents, referenced in Khurram's development work.18 documents
orchestration
Multi-agent coordination and workflow management; orchestrating autonomous agent interactions.15 documents
cloned
Created a copy or instance of AI agent or system with inherited context.8 documents
ambitions
Goals, desires, and identities of AI agent layer.7 documents
TASK
Semantic framework for structuring agentic instructions to Claude, critical for code generation accuracy.6 documents
agent actually
AI agent that demonstrates an end-to-end workflow pattern in his applications5 documents
agent architecture
Multi-agent system design with scheduled lead orchestration.5 documents
multi-agent
Orchestrating multiple AI agents working together coordinately.5 documents
agents running
AI agents actively executing or deployed in a system generating visible outputs.5 documents

Related

Seventeen years of working thought, indexed.
Ask me to present it in your conference room — WhatsApp +971 55 623 9111
Book Session →