Project Management Workflow Efficiency

You are paying for billions of tokens each day before generating a single useful output 💸 At Twitter, we cut ads ranking prediction costs by 85% - not with a better model, but by fixing payload bloat. The same pattern is showing up again with MCP. It’s brilliant for developer workflows, but naive production deployments create a “context-window tax” that compounds silently. Here's the math people aren't doing: → ~3,000 tokens of tool/schema context per request → 500k daily requests → billions of tokens/day Yes, caching helps - a lot. But only if prompts are structured for reuse. Most aren’t. Here are the top 4 things to solve this architecture problem: ❶ Default to cheap routers. Regex, embeddings, small fine-tuned models, or at most Flash/Haiku/nano-tier LLMs. Frontier models should be the last resort. The cost delta is 3–5x with negligible routing quality difference! ❷ Decouple orchestration from reasoning. Lightweight models handle tool use & APIs. Frontier models handle synthesis, multi-step reasoning, and ambiguity. Don’t use a sledgehammer to sort mail. ❸ Treat context like a production resource. Don’t inject every tool schema into every request. Scope tools, compress schemas, and load lazily. Every token costs on every call. ❹ Cache aggressively, but correctly. Prompt caching can cut costs up to 90% (Anthropic, OpenAI, Google DeepMind). But it only works if prefixes are stable and prompts are reusable. The best ML systems aren't the most clever. They're the ones that minimize tokens, isolate expensive reasoning, and make cost-quality tradeoffs explicit. This is Part 1 of my MCP production teardown. Over the next few weeks, I’ll share insights on Shadow AI protocols, model-agnosticism, memory vs reflex, and more. If you're building Gen AI systems at scale, I’d love to hear from you. Curious what’s been your highest cost or latency bottleneck so far.

Content marketing chaos looks like this: ➡️ CEO: can you write this ebook? ➡️ Product team: can I get a blog post series on this? ➡️ Customer success team: I have four customer stories for you to write THIS WEEK. This is no way to work. The writers get buried in ad hoc requests and end up scrambling to get work done. Details are slim, they're stressed, they spin wheels, you're wasting $, and there's no prioritization. The answer? A job request process. Most PM tools have form features you can use to collect information. Otherwise, there are standalone form tools you can use. In those forms you can ask questions like: • content type (drop down - blog, case study, ebook...?) • target audience (your specific buyer) • goal of content (awareness, book calls...) • how it will be used (email outreach? in a campaign?) • the customer question or problem it's answering 👎 Don't throw around terms like TOFU/BOFU and 'buyer persona' or make this resemble a content brief, because only a marketer can fill these out. Once you get the form: ✅ The requests can be reviewed, clarified, and assigned based on project details. ✅ You can start a back/forth to clarify or have a call if it requires a conversation. (And in some cases, the request goes away because it didn't make sense or there is a piece out there that will work). You can also create a PM board where people can just log and park ideas that can be explored later. No more ideas on post-it notes, mentioned offhandedly and forgotten, or buried in emails or Slack chats. This structured approach protects the team's time, allows for the right details to come to light, and results in content with purpose.

I discovered I was designing my AI tools backwards. Here’s an example. This was my newsletter processing chain : reading emails, calling a newsletter processor, extracting companies, & then adding them to the CRM. This involved four different steps, costing $3.69 for every thousand newsletters processed. Before: Newsletter Processing Chain (first image) Then I created a unified newsletter tool which combined everything using the Google Agent Development Kit, Google’s framework for building production grade AI agent tools : (second image) Why is the unified newsletter tool more complicated? It includes multiple actions in a single interface (process, search, extract, validate), implements state management that tracks usage patterns & caches results, has rate limiting built in, & produces structured JSON outputs with metadata instead of plain text. But here’s the counterintuitive part : despite being more complex internally, the unified tool is simpler for the LLM to use because it provides consistent, structured outputs that are easier to parse, even though those outputs are longer. To understand the impact, we ran tests of 30 iterations per test scenario. The results show the impact of the new architecture : (third image) We were able to reduce tokens by 41% (p=0.01, statistically significant), which translated linearly into cost savings. The success rate improved by 8% (p=0.03), & we were able to hit the cache 30% of the time, which is another cost savings. While individual tools produced shorter, “cleaner” responses, they forced the LLM to work harder parsing inconsistent formats. Structured, comprehensive outputs from unified tools enabled more efficient LLM processing, despite being longer. My workflow relied on dozens of specialized Ruby tools for email, research, & task management. Each tool had its own interface, error handling, & output format. By rolling them up into meta tools, the ultimate performance is better, & there’s tremendous cost savings. You can find the complete architecture on GitHub.

After optimizing costs for many AI systems, I've developed a systematic approach that consistently delivers cost reductions of 60-80%. Here's my playbook, in order of least to most effort: Step 1: Optimizing Inference Throughput Start here for the biggest wins with least effort. Enabling caching (LiteLLM (YC W23), Zilliz) and strategic batch processing can reduce costs by a lot with very little effort. I have seen teams cut costs by half simply by implementing caching and batching requests that don't require real-time results. Step 2: Maximizing Token Efficiency This can give you an additional 50% cost savings. Prompt engineering, automated compression (ScaleDown), and structured outputs can cut token usage without sacrificing quality. Small changes in how you craft prompts can lead to massive savings at scale. Step 3: Model Orchestration Use routers and cascades to send prompts to the cheapest and most effective model for that prompt (OpenRouter, Martian). Why use GPT-4 for simple classification when GPT-3.5 will do? Smart routing ensures you're not overpaying for intelligence you don't need. Step 4: Self-Hosting I only suggest self-hosting for teams at scale because of the complexities involved. This requires more technical investment upfront but pays dividends for high-volume applications. The key is tackling these layers systematically. Most teams jump straight to self-hosting or model switching, but the real savings come from optimizing throughput and token efficiency first. What's your experience with AI cost optimization?

The paradox of WIP limits contradicts every instinct about productivity. When demand increases, your natural response is taking on more work. Keep everyone busy. Maximize utilization. That's when delivery slows down. Little's Law explains why. Average cycle time equals work in progress divided by throughput. When WIP increases, cycle time increases proportionally. More items in the system means each item takes longer to complete. Context switching increases. Bottlenecks intensify. Quality issues emerge because nothing gets full attention. The solution feels counterintuitive. When pressure builds, lower your WIP limits instead of raising them. Fewer items in progress means each item moves faster. Faster movement means more completions. More completions reduce the backlog. The teams I work with resist this initially. Then they test it for two weeks. Cycle times drop by 30-40%. They never go back to the old way. Less work in progress creates more work completed. That's the paradox that accelerates delivery. #NavigateYourFlow #WIPLimits #LittlesLaw #FlowMetrics #Kanban

On average, it takes 8 days and 3.2 rounds of review to get a project deliverable approved. 7 in 10 project managers say chasing stakeholders for approvals slows down their teams significantly. This explains why projects fall behind schedule, resources are wasted, and deadlines become a constant source of stress. But… Because of these delays, project managers face constant roadblocks like: Endless email chains and follow-ups. Teams waiting idly for approvals that don’t come on time. Budget overruns caused by rework or missed timelines. Chasing approvals isn’t just time-consuming—it derails the entire project. When feedback or sign-offs are delayed, the ripple effect impacts everything: Planned resources go unused. Project milestones are missed. Team morale drops because of constant last-minute changes. Imagine this: You’ve coordinated with multiple stakeholders, only to spend days waiting for someone’s approval. Meanwhile: Your team is idle, wasting valuable hours. You’re scrambling to keep stakeholders aligned. Timelines are collapsing, and you’re stuck fixing the mess. This endless cycle of chasing approvals leaves you overwhelmed and exhausted. So, how do you take back control? The answer lies in streamlined approval workflows. Here are 3 actionable tips to get faster project approvals: 1. Set non-negotiable deadlines: Assign clear due dates for every review stage and automate reminders to keep stakeholders accountable. 2. Be specific in your requests: Specify exactly what needs to be approved whether it's a project charter, timeline, or deliverables so stakeholders know where to focus. 3. Centralize approvals: Use a single tool or platform for feedback and sign-offs to eliminate confusion and back-and-forth emails. The next time you’re stuck waiting for project approvals, ask yourself: • Have I communicated clear deadlines? • Am I specific about the feedback I need? • Is my approval process centralized and easy to follow? Take these steps, and you’ll not only stop chasing approvals but also keep your projects on track, under budget, and stress-free.

Helped a Hospital slash operational costs by 25% while improving patient care – here’s the breakdown A private hospital I worked with was facing two major problems: Rising operational costs eating into profit margins Declining patient satisfaction scores due to perceived cost-cutting They needed a way to reduce expenses without compromising care quality—or risk losing patients to competitors. 3 Strategic Changes We Made 1) Switched to Smart Inventory Management Reduced medical supply waste by tracking usage trends and automating reorders. Negotiated bulk purchase discounts with suppliers. 2) Optimized Energy & Infrastructure Costs Upgraded to energy-efficient lighting and HVAC systems. Shifted non-critical power usage to off-peak hours. 3) Reallocated Staff for Maximum Efficiency Cross-trained nurses and support staff to handle peak hours. Introduced telemedicine for minor follow-ups, freeing up doctors for critical cases. The Impact? ✅ 25% reduction in monthly operational costs ✅ 15% improvement in patient satisfaction scores ✅ Faster lab turnaround times due to streamlined workflows The best part? They maintained the same quality of care while saving ₹50+ lakhs annually—proving that cost optimization doesn’t mean cutting corners. Most hospitals think they have to choose between costs or quality, but the right strategy lets you improve both. If your hospital is struggling with high expenses or inefficient processes, DM me. Let’s find smart ways to boost your bottom line #healthcare #healthtech

I watched a client go from taking weeks to launch experiments... to literally a few hours. Here's how I helped them 👇 When I started working with this client, their experimentation program was slow. Every idea had to work its way through a maze of approvals, backlogs, and coordination across teams. By the time something launched, the window of opportunity had already closed. I wanted to help them move faster. MUCH faster. So I started introducing the things that set world-class experimentation cultures apart from everyone else. What, pray tell, might those things be? Well, I love explaining with a particular and very true story that goes around at a former employer: there was a copywriter who had an idea on her bike ride to work, came into the office (this is pre-pandemic, people), made the change in her CMS (which was connected to the experimentation platform), and launched it as a global experiment running before her first coffee. That’s the kind of speed I wanted my client to experience, and we made it happen. How? Sure, we tightened up their tech and data flows so experiments could run smoothly. But this was the minor point in all honesty. The real shifts came from bringing together a cross-functional team who had the skills to deliver autonomously, getting leadership backing for that team to take risks, and setting a clear and focussed goal for the team to rally behind. We removed unnecessary “approvals,” facilitated the essential conversations, created focus, and rewarded pace without compromising rigour (improving it, actually). The team became empowered to make their own decisions and built a culture that normalised risk-taking. The result was night and day. Just weeks earlier, ideas took months to get through approvals, builds, and launches. All before even monitoring, reporting, and decision making (if any) took place. Now? The team could go from an ideation session to launching quick wins in literally hours. And, I can tell you first hand, when a team experiences this shift from moving like a snail to sitting up front of a rocket ship, that acceleration brings creativity, confidence, and energy. The kind that spreads across teams and compounds. It’s infectious. Oh, and did I mention that they started getting more runs on the board too? 😉 So recap, what makes these cultures different? And what did we instil to help this team accelerate that fast? ✅ Leaders who empower and trust their teams. ✅ Teams with genuine ownership and motivation to create impact. ✅ A culture that celebrates learning, not just winning, and treats failure as fuel for improvement. That’s what lets someone move from an idea on a bike ride → to a global experiment before their first coffee. If you want to innovate and experiment at lightning speed, don’t just look at your tech. Look at your culture. Could your org handle that kind of speed? 👇

Break the work up. Watch the delivery speed up. Understanding how to break your work down into small independently shippable pieces of work is such a handy technique. Smaller pieces of work flow through your system faster. It's easier to have conversations in a code review about the work when it's only a couple of files. It allows you to evolve your design as you go instead of falling into a sunk cost fallacy around it or being exhausted by a litany of change requests. When you start a piece of work, take some time to understand all the individual components. Think about how you can sequence and deliver the work starting from the lowest levels until you reach those final high-level integration pieces. Even if you can't always break it all up, just thinking and chunking some of it will make for a far easier development process. There's so many benefits to working small, as an engineer, my favourite is the sense of momentum I get from seeing yet another PR deployed to production.

Struggling with Web Design Project Requirements? This One Framework Helps! Let’s face it – as a new web designer, tackling project requirements can be overwhelming. You’ve got client requests, user needs, and design goals to balance. If you're finding it tricky to connect all the dots, there's a simple framework that can change everything: (5W1H) This classic method (Who, What, When, Where, Why, and How) turns those confusing project requirements into a clear, manageable checklist. Here’s how it can help: ✔ Who Identify the end-users. Who is the website for? Understanding their needs ensures every design choice connects with the audience. ✔ What Define what the website aims to achieve. Is it selling a product, building a brand, or sharing information? This guides everything from layout to features. ✔ When Consider any timelines or specific events. Is it for a campaign launch or seasonal sale? Knowing this helps you prioritize features and updates. ✔ Where Think about where the site will be accessed – desktop, mobile, tablet? This affects layout, font sizes, and user experience. ✔ Why Ask why the client wants this project. Knowing the motivation clarifies the purpose, giving you a strong foundation for creative decisions. ✔ How Finally, think about how users will interact with the website. This impacts navigation, CTAs, and design flow. 💡 Using 5W1H might seem simple, but it brings much-needed clarity, helping you decode project requirements and build websites that truly resonate. Ready to try it in your next project?