Prioritizing Outcomes vs. Methods in LLM Projects

Start With the Use Case, Not the LLM When I first began guiding teams through Gen AI adoption, most conversations started with: “Which LLM should we use: Cohere, OpenAI, Anthropic, or Gemini, etc?” That’s a trap I’ve seen even fall into. The real question should be: “What business outcome are we solving for?” An LLM-first mindset leads to tool sprawl, multiple APIs, overlapping features, and no measurable ROI. A use-case-first approach, on the other hand, forces clarity: What process are we improving? What knowledge or data powers it? How will success be measured? Only once those questions are clear does the LLM platform matter, and often, you’ll find the answer isn’t a single model, but a combination of tools that best fit the workflow. A Simple Framework Define the Outcome: Start with a business metric, time saved, revenue increased, risk reduced. Identify the Friction Point: What’s slowing this process down? Data? Human effort? Latency? Match Platform Capabilities: Choose the AI stack (e.g., LLM, vector DB, agentic tools) that targets that friction. Prototype Fast, Measure Early: Build a thin slice of value before expanding.

A week ago I was talking with a VP of AI, a colleague from my network. We ended up on a familiar topic: why so many AI projects stall. The conclusion was that it’s not usually the models. It’s the lack of prioritisation. Too often, someone jumps straight into the “hottest” LLM or agent idea. A few months later? Sunk costs, scattered pilots, no ROI. I’ve seen this pattern across companies. At one workshop, a global tech leader brought together 50+ engineers and business leaders. They generated 150+ ideas for LLM features. After scoring for feasibility and impact, only two made the cut, and those two were the ones that aligned with ROI and scaling. Another client arrived with a long wish list of AI projects across departments. After structured evaluation, they left with one clear initiative. Not the flashiest, but the one that delivered immediate productivity gains and could scale without drama. That’s the real inflection point: when technical depth meets business value. Get it right → you move fast, scale confidently, and win buy-in. Get it wrong → you spend months debugging projects that should never have started. Over the years, in different companies and finally at deepsense.ai, I’ve learned the same lesson: tech projects don’t fail because LLMs are weak. They fail when leaders skip prioritisation. This is becoming increasingly relevant, as starting a project is becoming easier, but launching them successfully is becoming harder. What do you think?

🚀 Some observations from our efforts to build Applied AI products at scale The past few months have been a masterclass for me and my team in learning how to build and ship Applied AI products in environments where both the tech and the business move at very different speeds and at Uber scale. A few takeaways: 1️⃣ The ecosystem is "very" fragmented and evolving fast. What looks like a solid choice today may feel outdated a month later. In large organizations, where decision cycles are long, this can get messy quickly. The way forward: make decisions at the right level of abstraction—anchored to the business problem, not just today’s tech capabilities—so you can adapt, mix, and evolve over time. 2️⃣ Don’t get lost in the vendor / model rabbit hole. You can spend months mapping who does what, with what model and end up with a pretty Gartner-style chart and some amazing decks…but solve zero problems. Keep the focus on the business problems first, make tech choices expendable, and once you have a viable solution—ship it. There’s no “perfect” solution out there, the same models offer different results with a lot of fine tuning and as long as you're making progress on your key metrics, you're on the right path. 3️⃣ Leaders need a framework to cut through the noise. When explaining AI options, anchor on a simple set of dimensions: Quality Outcomes Cost Feature depth Latencies The last one is underrated: most LLMs are too slow for true real-time decisions at the quality you need. Make it faster and hold the quality bar and the cost blows up 10x. That’s why augmenting with other ML models is essential. 4️⃣ Define what “good” means for you. This isn’t just about picking the right model—it’s about building the internal muscle: independent datasets, human & automated evaluations, publishing benchmarks, and understanding how good an AI system is in the context of your business problem. The tradeoff matrix of Quality of outcomes vs Cost is where the real decisioning happens. The AI world is moving crazy fast, but anchoring on business problems, building pragmatic frameworks, and defining your own defensible “good” is how you cut through the chaos. cc: Srinivas Motamarri, Gourav Gupta, Pooja Garg, Ryan Ashby, Saiket Talukdar, Ashish Krishna V., Javed Abdulla, janardhan reddy, Suchitha C.

𝗔𝗿𝗲 𝘄𝗲 𝗱𝗲𝗹𝗶𝘃𝗲𝗿𝗶𝗻𝗴 𝗿𝗲𝗮𝗹 𝘃𝗮𝗹𝘂𝗲 𝗼𝗿 𝗷𝘂𝘀𝘁 𝘀𝘁𝗮𝘆𝗶𝗻𝗴 𝗯𝘂𝘀𝘆? 𝗔 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝘁𝗵𝗲 𝗮𝗴𝗶𝗹𝗲 𝗰𝗼𝗺𝗺𝘂𝗻𝗶𝘁𝘆 𝘀𝗵𝗼𝘂𝗹𝗱 𝗿𝗲𝗳𝗹𝗲𝗰𝘁 𝗼𝗻 𝗺𝗼𝗿𝗲 𝗼𝗳𝘁𝗲𝗻.   Too frequently, teams and organizations celebrate their output: The number of completed stories, the amount of shipped features, the velocity achieved in a sprint. These are tangible and measurable indicators – and thus often mistaken for real progress.   But are they?   ➡️ 𝗢𝘂𝘁𝗽𝘂𝘁 𝗶𝘀 𝘄𝗵𝗮𝘁 𝘄𝗲 𝗽𝗿𝗼𝗱𝘂𝗰𝗲; 𝘁𝗵𝗲 𝗶𝗺𝗺𝗲𝗱𝗶𝗮𝘁𝗲, 𝘃𝗶𝘀𝗶𝗯𝗹𝗲 𝗿𝗲𝘀𝘂𝗹𝘁 𝗼𝗳 𝗼𝘂𝗿 𝘄𝗼𝗿𝗸.   ➡️ 𝗢𝘂𝘁𝗰𝗼𝗺𝗲, 𝗼𝗻 𝘁𝗵𝗲 𝗼𝘁𝗵𝗲𝗿 𝗵𝗮𝗻𝗱, 𝗶𝘀 𝘄𝗵𝗮𝘁 𝗼𝘂𝗿 𝘄𝗼𝗿𝗸 𝗮𝗰𝗵𝗶𝗲𝘃𝗲𝘀; 𝘁𝗵𝗲 𝗮𝗰𝘁𝘂𝗮𝗹 𝗶𝗺𝗽𝗮𝗰𝘁 𝘄𝗲 𝗰𝗿𝗲𝗮𝘁𝗲 𝗳𝗼𝗿 𝘂𝘀𝗲𝗿𝘀, 𝗰𝘂𝘀𝘁𝗼𝗺𝗲𝗿𝘀, 𝗼𝗿 𝘁𝗵𝗲 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀.   That’s where the real difference lies.   Three practical examples:   ➡️ 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 𝗶𝗻 𝗘-𝗖𝗼𝗺𝗺𝗲𝗿𝗰𝗲 - Output: A new filter feature is released on the online store. - Outcome: Bounce rates on product pages drop by 20%, and conversion rates improve significantly because users find what they’re looking for faster.   ➡️ 𝗜𝗧 𝗦𝗲𝗿𝘃𝗶𝗰𝗲 𝗶𝗻 𝗮 𝗙𝗶𝗻𝗮𝗻𝗰𝗶𝗮𝗹 𝗜𝗻𝘀𝘁𝗶𝘁𝘂𝘁𝗶𝗼𝗻 - Output: A new internal ticketing system is rolled out and documented. - Outcome: Average response times decrease from five to two days and satisfaction across business departments rises noticeably.   ➡️ 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 & 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 𝗶𝗻 𝗮 𝗖𝗼𝗿𝗽𝗼𝗿𝗮𝘁𝗶𝗼𝗻 - Output: A comprehensive e-learning module on leadership is developed. - Outcome: Managers apply new techniques in their daily work, resulting in a measurable increase in employee engagement within their teams.Outcome means impact. Benefit. Change. It answers the question: What has truly improved because of our work?   In the agile space, we often speak of “delivering value early and often.” But too often, we equate success with speed, efficiency, or throughput instead of measurable and validated value.   What we need is a fundamental shift in focus: ⏺️ From completing backlogs to validating outcomes ⏺️ From feature delivery to behavioral impact ⏺️ From busy teams to learning organizations   Outcome-driven work means: ⏺️ Truly putting customers at the center ⏺️ Measuring success by value, not by volume ⏺️ Making learning a core element of the development process   This is more than a mindset shift it's a cultural transformation. Because when we prioritize outcomes over output, we don’t just change how products are built. We change how organizations think, decide, and succeed. Agility is not about moving faster.   𝗜𝘁’𝘀 𝗮𝗯𝗼𝘂𝘁 𝘄𝗼𝗿𝗸𝗶𝗻𝗴 𝘀𝗺𝗮𝗿𝘁𝗲𝗿 𝘄𝗶𝘁𝗵 𝗽𝘂𝗿𝗽𝗼𝘀𝗲, 𝗶𝗻𝘁𝗲𝗻𝘁, 𝗮𝗻𝗱 𝗺𝗲𝗮𝘀𝘂𝗿𝗮𝗯𝗹𝗲 𝗶𝗺𝗽𝗮𝗰𝘁.   #agile #transformation #management

"Focus on outcomes vs. outputs. Features don't automatically create value" The pivot from an emphasis on outputs to outcomes can be a critical paradigm shift for teams intent on building solutions that provide real value and enhance customer satisfaction. I just wrote about this topic in my new article on LeadDev: https://lnkd.in/gYqwH32S An output signifies the result of an activity (for example, launching a new feature), whereas an outcome is a change in customer behavior that drives business results (e.g. user-happiness improved, thanks to this over time we may see an impact to business metrics like sales). Real-world Scenario Imagine you're Spotify. A newly launched feature (an output), enables listeners to swiftly save their favorite songs (an outcome). The outcome of this results in heightened user satisfaction with usability, an increase in subscription numbers, and a subsequent rise in sales over the following six months (impact). Guiding Teams: Outputs, Impacts, and Outcomes To optimize efficacy, it's imperative to understand the distinctions between guidance based on outputs, impacts, and outcomes: Guidance based on Outputs: This entails asking teams to develop specific new features, products, or enhancements. However, this approach does not inherently ensure the provision of value to the users or the business. Guidance based on Impact: This encourages the team to deliver overarching value, such as revenue enhancement or cost reduction. Although essential, this may not offer explicit direction for the team or assist in the creation of user-oriented solutions. Guidance based on Outcomes: This involves asking teams to create specific customer behavior changes that drive business results. This allows teams to find the right solution and keeps them focused on delivering value to the users and the business. Employing an Outcome-Oriented Approach The value of an outcome-oriented approach is evident when uncertainties abound in a new initiative, product, or feature. Such uncertainties are a common occurrence in software engineering, and outcomes offer a means to define goals that encourage teams to experiment with various solutions until the right one is identified. The Limitations of an Outcome-Oriented Approach In instances where a solution is almost guaranteed to work, such as routine maintenance or bug fixes, the outcome-oriented approach may not be the best fit. For these scenarios, an output-focused plan is more fitting. Occasionally, I'm asked about where elements with no clear link to outcome fit in (like technical debt, code health). This often boils down to perspective (e.g., neglecting these could increase the long-term cost of outcome execution, framing in terms of the value back to the business, but can still have their place). Concentrate on what truly counts. Illustration credit: Workpath: https://lnkd.in/gFRcs-v7 #softwareengineering #motivation #productivity #lifeatgoogle

Have you heard: "focus on outcomes, not outputs"? If you're a product manager, there is a high chance you have. This is good advice, but it is incomplete For the following reasons: 1. Some PMs don't understand 𝘸𝘩𝘢𝘵 are outcomes 2. Some don't know 𝘸𝘩𝘺 it is imp. to focus on outcomes 3. Some might not know 𝘩𝘰𝘸 to be outcome driven Let's answer all three questions: 𝗪𝗵𝗮𝘁 𝗮𝗿𝗲 𝗼𝘂𝘁𝗰𝗼𝗺𝗲𝘀 (𝗮𝗻𝗱 𝗵𝗼𝘄 𝗮𝗿𝗲 𝘁𝗵𝗲𝘆 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗳𝗿𝗼𝗺 𝗼𝘂𝘁𝗽𝘂𝘁𝘀) Outputs are the deliverables or products that a team produces Outcomes refer to the impact or the value that these outputs have on the customers and the business. Example: you're a PM for a food delivery app;  responsible for increasing retention. You're launching a feature that will increase retention by 5% 👉 If you launch the feature on time without issues, that is an OUTPUT 👉 If you ensure that the retention increases by 5% either through the same feature or any other feature, that is an OUTCOME. 𝗪𝗵𝘆 𝘀𝗵𝗼𝘂𝗹𝗱 𝘆𝗼𝘂 𝗳𝗼𝗰𝘂𝘀 𝗼𝗻 𝗼𝘂𝘁𝗰𝗼𝗺𝗲𝘀 1. Prioritizing outcomes over outputs shifts the focus from just delivering features to delivering value. 2. It ensures you align your efforts with the overall business goals 3. That enables your products to create the desired impact 4. Alignment between product and business goals makes decision making and prioritisation more impactful 𝗛𝗼𝘄 𝘁𝗼 𝗳𝗼𝗰𝘂𝘀 𝗼𝗻 𝗼𝘂𝘁𝗰𝗼𝗺𝗲𝘀 (𝗼𝘃𝗲𝗿 𝗼𝘂𝘁𝗽𝘂𝘁𝘀) 1. 𝗦𝗲𝘁𝘁𝗶𝗻𝗴 𝘁𝗵𝗲 𝗿𝗶𝗴𝗵𝘁 𝗴𝗼𝗮𝗹𝘀 Set the right goals -- goals that are tied to value. Then, do whatever it takes to achieve these goals. 2. 𝗜𝗺𝗽𝗹𝗲𝗺𝗲𝗻𝘁𝗶𝗻𝗴 𝗢𝘂𝘁𝗰𝗼𝗺𝗲-𝗕𝗮𝘀𝗲𝗱 𝗥𝗼𝗮𝗱𝗺𝗮𝗽𝘀 Focus your roadmap on impact that the product aims to achieve. This ensures that the team prioritises value over on-time delivery 3. 𝗧𝗿𝗮𝗻𝘀𝗽𝗮𝗿𝗲𝗻𝗰𝘆 𝗮𝗻𝗱 𝗽𝗿𝗼𝗮𝗰𝘁𝗶𝘃𝗲𝗻𝗲𝘀𝘀 Proactively and repeatedly inform others about your outcome-based goals. That will make expedite collaboration and make it easier to achieve the outcomes. 4. 𝗔𝗹𝘄𝗮𝘆𝘀 𝘁𝗵𝗶𝗻𝗸 𝗮𝗯𝗼𝘂𝘁 𝘁𝗵𝗲 𝘂𝘀𝗲𝗿 All effort, resources, and bandwidth the team spends should ALWAYS lead to net positive value creation for the user. Every 3-6 months: stop whatever you're doing, and ask "will these things create more value for the user?" Yes? Let's keep investing in it. No? Stop and focus on other things. 𝗧𝗵𝗮𝘁 𝗱𝗼𝗲𝘀𝗻'𝘁 𝗺𝗲𝗮𝗻 𝗳𝗼𝗰𝘂𝘀𝗶𝗻𝗴 𝗼𝗻 𝗼𝘂𝘁𝗽𝘂𝘁𝘀 𝗶𝘀 𝗯𝗮𝗱 Outputs are a required prerequisite for creating impactful outcomes. The timely delivery of features and products is essential for achieving the desired impact on users and the business. A good output-outcome balance ensures that the pursuit of driving value is grounded in a practical approach that considers the delivery and execution of products and features. -- That is it. A short, but an important guide to help PMs understand the meaning and importance of outcomes Do you focus on outcomes or outputs?

I am currently reviewing chapters for our forthcoming book about the use of GenAI for qualitative analysis (SAGE), co-edited with David Morgan. I loved the approach by Jonas Wibowo and Hendrik Wiese because they did not immediately give up on or reject the new technology when something didn’t work. They recognized the weaknesses of LLMs and adapted their approach to utilize their strengths. The general aim of their analysis was to reconstruct themes related to the research questions. They stayed within familiar methodological territory, using categorical thinking as their analytic strategy (e.g., Freeman 2017). So yes—their goal was to build a hierarchical category frame. Dominique Ruggieri, PhD Allison E. Curry, PhD, MPH Ryan Warren Regarding our discussion of how to bridge the old with the new: this example shows that we can stay conceptually with what we have always done yet change the procedures. Coding data does not just mean “tagging”—it can be implemented differently. I truly believe this is the direction we need to aim for. Andrew Katz Matthew Nyaaba But let’s take a look at what Jonas and Hendrik have done. They discovered that the LLM could not properly distinguish between different hierarchical levels; it was inconsistent and suggested different labels when asked again—all the known pitfalls described elsewhere. Others might have concluded that LLMs are not useful because they aren’t reliable, nuances are lost, and results can’t be reproduced. Instead, they stopped asking the LLM to do what it cannot do well and embraced its nature—its variability, its capacity to generate rather than produce final outcomes (this is what we are essentially trying to do when we "force" it to tag data). The human researcher remained in charge of building the hierarchical scaffolding for the project, while the LLM provided inspiration and plausibility checks, pulled out supporting evidence from the data, and tested the suggested structure. The AI assistant offered pattern ideas; the researcher collaborated with it to elaborate patterns, request supporting cases, retrieve data, and jointly validate interpretations to make sense of the data. They found that meaning-making suffered whenever too much of the process was handed over to the AI. The human researcher proved essential in the interpretive work. They concluded: We succeeded by using the strengths of the GAI while ignoring its weaknesses. I would add—put the human researcher in charge where the LLM fails and keep the AI as a creative partner, never the final decision-maker. Isn’t that what building an effective team is all about? Identifying each person’s strengths and assigning tasks they’re great at—instead of setting them up to fail by giving them the wrong responsibilities, only to criticize them afterward. Kien Nguyen-Trung, PhD Dr. Piyush Gotise #LMMStrenght #AIQualitativeAnalysis #QInsights #QualitativeMethods #QualitativeResearch #TaggingVsCoding #MethodologyVsMethods

I've spent the past year collaborating with teams from various industries to develop LLM agents. Through this experience, I've gathered valuable insights: - Simple workflows outperform complex agents. - Prioritizing cost and speed is more effective than focusing on extravagant features. - Many tasks can be accomplished without the need for full AI agents. Key components that matter most: - Efficient task routing - Basic evaluations system - Clear and concise prompts - Thoughtful coordination Simplify your LLM systems: - Emphasize transparency - Prioritize efficiency - Focus on practicality By following these principles, you can create systems that yield tangible outcomes. For further reading, check out Anthropic's insightful blog on constructing efficient AI agents – a go-to resource in my arsenal for guiding others on AI implementation strategies. Link to the blog in the comments section. #ai #LLMs #agent #frameworks

3 secrets behind 100+ successful Quote-to-Cash projects (that resulted in 100+ million in customer revenue): 1. Focus on outcomes, not just requirements. Share the desired business outcomes during the sales process, not just the technical requirements. Why? Because the project champion buys the software based on expected results, but the vendor only focuses on the "feature" or the "requirement”. This misalignment often leads to unmet expectations and disappointed clients. By prioritizing outcomes: - You ensure client satisfaction - You deliver the expected growth and results It's a win-win for the client and the vendor. 2. Start with an MVP (Minimum Viable Product). An MVP helps achieve quicker time to value. A ten-month timeline? Not a chance. Aim for ten weeks. This way, you start realizing value sooner. For example: If the desired outcome is to implement software to give your reps 30% of their time back so they can focus on selling and crush their quota, start with an MVP that achieves 15% back. Then follow up with other phases that get the rest of the outcome. (reclaiming up to 30% of your time) And shorter projects face fewer changes, making success more achievable. 3. Success breeds success. There will always be naysayers, so early wins with an MVP help align expectations and gain support from stakeholders. The best part? You'll turn skeptics into supporters (and that's a great feeling). To recap: Focus on outcomes. Embrace the MVP approach. Leverage early wins to build momentum. Which 'secret' resonated most with you - 1, 2, or 3? Let me know. P.S. Follow me for more content like this.

"We need to prioritize our roadmap, but every stakeholder has a different opinion." The problem isn't conflicting opinions—it's the lack of objective criteria for evaluation. Traditional prioritization methods that fail: - Executive opinions and gut feelings - Revenue projections based on assumptions - Competitive feature comparisons - Engineering complexity assessments - Sales team requests and customer demands Why they fail: None directly measure potential to create customer value. The Outcome-Driven alternative: Step 1: Evaluate each initiative against underserved customer outcomes Step 2: Score based on ability to address high-opportunity areas Step 3: Consider cost, effort, and risk factors Step 4: Optimize high-value projects for maximum impact The difference: Instead of guessing which projects will succeed, you're investing in solutions that address known customer outcomes. Companies using this approach achieve 86% success rates versus the industry average of 17%. The question isn't whether you should prioritize your pipeline—it's whether you're using the right criteria. What would change if every project decision was based on customer outcome data?