Insights on enterprise AI
Practical guidance, case studies, and analysis on AI strategy, RAG, agents, automation, and integration — written by IDS engineers and consultants.
From RAG to GraphRAG: when vector search isn’t enough for legal, finance, and engineering docs
Vector search finds chunks similar to your query — that’s the whole mechanism. For legal contracts, financial filings, and engineering BoMs where relationships between entities matter, similarity isn’t structure. A practical guide to GraphRAG and the hybrid retrieval pattern that fits most enterprises.
Knowledge graphs + LLMs for Vietnamese enterprises: handling language nuance at scale
Vietnamese tone marks. Compound-noun word boundaries. Company-name conventions (Công ty Cổ phần / TNHH / JSC). Administrative restructuring of districts and wards. Code-switching with English. Regional vocabulary. Six realities that break off-the-shelf retrieval — and how a knowledge-graph layer handles them.
Computer-use agents vs. legacy RPA: where each one actually belongs
UiPath and Automation Anywhere aren’t dead — they’re still doing real work in real enterprises. But computer-use agents handle the tasks RPA was always bad at. A six-question allocation rule and the hybrid pattern most enterprises actually need.
Beyond chatbots: agentic AI is finally crossing into core enterprise workflows
Agentic AI — models that plan, call tools, verify their own outputs — has crossed the threshold from demo to production. Three things change in the architecture, three workflows earn it first, and one rule of thumb tells you when not to reach for an agent.
Voice AI for Vietnamese customer service: dialects, code-switching, and brand voice
Voice AI works well in English. For Vietnamese customer service, off-the-shelf stacks miss three things — regional dialect variation, English/Vietnamese code-switching mid-call, and brand-appropriate Vietnamese register. Each one shows up in CSAT before the engineering team notices.
The token economics of scale: keeping AI costs flat as usage 10×s
Token cost grows linearly with usage. Five well-known levers — model routing, prompt caching, response budgets, batch APIs, eval-driven downgrades — compound to flatten that curve. Most teams pull them out of order. The eval suite is the prerequisite for the biggest savings.
The CFO's AI scorecard: measuring real ROI in the first 12 months
Most AI projects fail the CFO test not because they didn’t work but because nobody measured them in finance terms. Four buckets — revenue, cost, risk, capability — each with a baseline, a target, and a 30/60/90 cadence so the answer in month twelve doesn’t rest on storytelling.
Building an LLM threat model: a 7-step framework for enterprise AI
STRIDE doesn’t fit. OWASP’s LLM Top 10 is a taxonomy, not a process. Compliance checklists ask the right questions for the wrong systems. A seven-step framework that produces a CISO-signable artifact and a runbook your engineering team will actually use.
Beyond prompt injection: data exfiltration risks in enterprise AI agents
Prompt injection is the entry point. The interesting question is what the agent does next. Four exfiltration patterns appear repeatedly in real enterprise AI agent deployments — each one has an architectural remediation, not a prompt-level one.
Showing 9 of 20 articles.