About
Are your transformation initiatives stalling? Are you under pressure to deliver at scale,…
Articles by Steve
Activity
13K followers
-
Steve Pereira reposted thisSteve Pereira reposted thisHow hard should you work to solve a problem? Maybe hard. Maybe not—if you eliminate it altogether. Like processes. Why spend time improving a process that shouldn't exist in the first place? Like reporting. Which, in many (most?) organizations, is a mess. During a client assessment this week, we uncovered a weekly report that was robbing leaders of valuable time. While it *only* takes 10 minutes to complete (best case process time), that's 2,383 hrs per year across the organization. ➡️ That's the equivalent of paying more than one leader a full salary to do nothing but complete a report. And the report? - Requires information that already exists in systems - Asks for monthly data . . . in a weekly report - Includes questions where the answers are already known The solution is NOT: - Automating the report - Using AI to complete the report The report—in its current form—needs to go. Some of the information IS vital to know weekly. But it needs to be delivered in a more efficient way, with less human effort. Those 2,383 hours (freed capacity) could be spent: - Coaching and developing teams - Solving real problems - Spending time with customers - Planning for growth - Not working 60 hours a week And here's the part most organizations miss: Reports don't deliver value to customers unless they actually use them. If you want to clean this up, start here: 1️⃣ Assemble a cross-functional team (no more than 10) 2️⃣ Decide who needs what information and when (cadence)—with a bias toward data that drives decisions, not "nice to know" 3️⃣ Conduct a current state "report inventory": who gets it, when, and why. 4️⃣ Revisit actual needs based on what you find. 5️⃣ Redesign how information is delivered (format matters!)—with as little human effort as possible. Question every report. Quantify the time. Eliminate what shouldn't exist. Cleaning up reporting is one of the most respectful things you can do for your people. And it will almost always reveal bigger problems worth solving.
-
Steve Pereira reposted thisSteve Pereira reposted thisHow to Use Value Stream Maps to Reinforce Agility & Effectiveness, Part 1 https://lnkd.in/e6qqPJGv #Teams #planningHow to Use Value Stream Maps to Reinforce Agility & Effectiveness, Part 1 - Johanna RothmanHow to Use Value Stream Maps to Reinforce Agility & Effectiveness, Part 1 - Johanna Rothman
-
Steve Pereira shared thisJust found this gem of a screenshot and had to share it! One of my earliest experiences with Flow Engineering in a fast-moving agency. Fast-moving, but struggling to translate that motion into progress and value. We initiated outcome mapping to align on a single target, and the voting results revealed a surprising outcome: every participant had a different target outcome in mind. This was an incredible signal that highlighted the underlying misalignment within the team, the reason so much was happening but little was getting done. In the following moments, we transformed that confusion into clarity by aligning on a common target. From that point forward, we made significant progress together. It raises an important question to reflect on with your team: Are we all aligned on what the target is? #flowengineering
-
Steve Pereira shared thisIs "Human in the loop" viable in 2026? It could be your biggest constraint right now. What "Human in the loop" always attempts to do is inspect quality into an output instead of building quality in. What is that human doing? What of that can't be defined in policy? Guidance? Code? An effective, scalable, high-throughput system of work looks a lot like the internet. There's no human inspecting the packets - the work is delegated to tools that know what violates policy, and the humans define policy. The internet is such an effective and efficient system of work that those policies can actually be exceedingly simple. (the simplicity is also why it's so effective and efficient) The next systems of work will look a lot like that. Work will come from anywhere and go anywhere that can facilitate value delivery. Humans define and shape policy, and the system is engineered to align to it. Maybe that sounds a bit inhuman, but is our highest purpose to act as police and judges? Should we spend our time checking, or creating? Maybe "Human in the loop" is no longer viable, but was it ever? Source: Wiring for Flow, 2024
-
Steve Pereira reposted thisSteve Pereira reposted thisAxiom just announced that Flow Engineering won the Gold Medal in their 2026 Business Reference book award. Founded in 2007, the Axiom Business Book Awards recognize outstanding new titles that drive professional growth and inspire positive change. As business is speeding up, the ability to take a high-level perspective on how to help teams work better together keeps getting more important. Check the link in the comments to download a free chapter from the book. Axiom award announcement: https://lnkd.in/esDDFwuw)
-
Steve Pereira shared thisAutoresearch and increasingly autonomous agents are bringing us closer and closer to mechanisms we've struggled to grasp and implement for decades - and may never have effectively adopted. While 1st order cybernetics was about observed systems (thermostats) and 2nd order was about observing systems (a therapist observing couple dynamics), 3rd order cybernetics brings us into co-evolved systems (many participants shaping policy). It gets fuzzier but more powerful and consequential at higher levels. 3rd order cybernetics (learning about and improving a system of learning) is extremely difficult for humans. We struggle to connect cause and effect, grapple with scope and scale, detect weak signals and anomalies, and notice patterns in immense data sets over time. We have a persistent habit to 'try harder next time'. AI systems were at 'try harder next time' in 2023, and they won't ever be back. We struggle with failure. AI systems are numb to it. Of course we can train ourselves to conquer these things: Empirical thinking, relentless iteration - but does it make us more human? Automated systems can mathematically and conclusively converge and maintain integrity such that they can always build on what works to arrive at 'better', and rebuild at any time without a 'sunk cost' or other biases we hold tight. This may sound terrible for humans - we're so bad at so many of these behaviors - but I don't think we were ever meant for these things. We're not built for recursion, iteration, predictability, scale, reliability - pick any 'ility'! We're meant for creativity, passion, exploration, discovery, curiosity. We spend far too much time in the 1st order, and not nearly a fraction of a sliver of enough at the 3rd. To even grasp the 3rd we need new tools built and enabled by machines that don't yet exist. Handing off the mechanical to machines is a path to much, much more humanity, and less of what I would gladly have less of. We have so much to explore and imagine!Let It Crash: How Erlang, Wiener, and Ohno Taught Me to Build AI Agents That LearnLet It Crash: How Erlang, Wiener, and Ohno Taught Me to Build AI Agents That Learn
-
Steve Pereira reposted thisSteve Pereira reposted thisOn reflection, I can see why AI looks amazing to folks who: - Didn't automate stuff before - Didn't start by writing a test - Didn't reflect on their deployment pipeline and remove waterfall steps - Didn't merge to main every hour or two But, oh my, there are other things you're still missing out on, like feedback loops, that are going to cause you harm. I find XP/CD practitioners are largely unimpressed as they've already achieved the throughout (and also maintained high stability). Compared to "code as fast as you can" folks, I can imagine XP/CD folks using this technology in a very different way, with very different goals, if at all.
-
Steve Pereira reposted thisSteve Pereira reposted this"The coding bottleneck is dead. Long live the new bottleneck." AI coding agents are dissolving what used to be a bottleneck in the product development factory. Teams are building in hours what used to take weeks. The productivity gains are real — I'm seeing it firsthand and hearing about it from my clients, even in real-world large-scale companies. But here's what keeps coming up when exploring the impact of AI for coding and knowledge work with tech, product, and AI strategy leaders: What is the next bottleneck? "an hour lost at a bottleneck is an hour out of the entire system. an hour saved at a non-bottleneck is worthless." (Goldratt - Theory of Constraints) A reality where we build faster than ever is on the horizon, or even here for some teams. But coding/building hasn't really been the bottleneck for a while... The bottleneck, even before AI accelerated coding, has been Value and Validation... Value — are we building the right things? Aligned with strategic needs? Looking across value streams holistically? This was always hard. Delegating to AI doesn't magically solve it. Validation — not just "does it work?" but "is it useful to its customers?" You can ask AI agents to loop and iterate until they meet a strong definition of done — even one that satisfies GxP auditors. But will what they build be valuable? That's where AI hits the same wall humans have always hit: closing a real feedback loop. How long does it take you from deploying a product to knowing for sure whether it's useful? Unless you're a B2C tech giant, your agentic-enabled coders might be 10x faster, but will still have to wait at the tollgate to get real feedback. There's actually a risk of the same AI slop we see here on linkedin becoming more and more product features that work, but aren't useful. Effective product leadership - clarity of outcomes, sufficient context, enabling aligned autonomy - is a good start for limiting the AI-coded slop. Figuring out how to 10x the speed of learning, not just the speed of building - that's next frontier. figuring out useful leading indicators that can be observed in tight feedback loops. figuring out observability in environments where it's more than just installing datadog. Creating agentic "ralph wiggum loops" that go beyond a "quality check" to "usefulness check". #AI #Strategy #Agentic #ProductThinking #Value #Validation
-
Steve Pereira reposted thisSteve Pereira reposted thisMost AI initiatives don't fail at implementation. They fail before a single prompt is written. On April 21st, Josh Burnley is joining Steve Pereira (CEO, Visible Flow Consulting) for a candid executive conversation on what actually gets in the way and what to fix first. 11 AM CT. Register here: https://lnkd.in/gkdpNACR
Steve Pereira commented on a post
1d
This is such an important status quo to challenge in enterprises!! I see so many ‘business intelligence’ departments without question: what decision is being made, and by what method is it made most effectively and efficiently?
There is so much opportunity to bring a recommendation, with context, to a decision maker when a decision must be made - but there’s an even greater opportunity to question why that decision can’t be made once and built into code!
Steve Pereira replied to a comment
2d
Jared Hill under any circumstances; time is a constant, we can only do one thing at a time, outputs need inputs, we need to know what is happening, and measure our efforts, so I imagine Kanban works under any circumstances
-
Steve Pereira reacted on thisSteve Pereira reacted on thisHow hard should you work to solve a problem? Maybe hard. Maybe not—if you eliminate it altogether. Like processes. Why spend time improving a process that shouldn't exist in the first place? Like reporting. Which, in many (most?) organizations, is a mess. During a client assessment this week, we uncovered a weekly report that was robbing leaders of valuable time. While it *only* takes 10 minutes to complete (best case process time), that's 2,383 hrs per year across the organization. ➡️ That's the equivalent of paying more than one leader a full salary to do nothing but complete a report. And the report? - Requires information that already exists in systems - Asks for monthly data . . . in a weekly report - Includes questions where the answers are already known The solution is NOT: - Automating the report - Using AI to complete the report The report—in its current form—needs to go. Some of the information IS vital to know weekly. But it needs to be delivered in a more efficient way, with less human effort. Those 2,383 hours (freed capacity) could be spent: - Coaching and developing teams - Solving real problems - Spending time with customers - Planning for growth - Not working 60 hours a week And here's the part most organizations miss: Reports don't deliver value to customers unless they actually use them. If you want to clean this up, start here: 1️⃣ Assemble a cross-functional team (no more than 10) 2️⃣ Decide who needs what information and when (cadence)—with a bias toward data that drives decisions, not "nice to know" 3️⃣ Conduct a current state "report inventory": who gets it, when, and why. 4️⃣ Revisit actual needs based on what you find. 5️⃣ Redesign how information is delivered (format matters!)—with as little human effort as possible. Question every report. Quantify the time. Eliminate what shouldn't exist. Cleaning up reporting is one of the most respectful things you can do for your people. And it will almost always reveal bigger problems worth solving.
-
Steve Pereira liked thisSteve Pereira liked thisCode review got automated. Testing got automated. Analysis got automated. Requirements review? Still a human conversation in a 30-minute refinement meeting. That was fine when developers were the ones interpreting specs. Humans fill gaps. They ask follow-up questions. They pattern-match from context that never made it into the ticket. AI coding tools don’t do any of that. They read the requirements literally and build. Ambiguity becomes architecture. Missing context becomes missing functionality. The output is only as good as the input. Every developer experience survey says the same thing: incomplete requirements are the top frustration. Now those same incomplete requirements are the primary input for AI-generated code. The gap between what’s in the ticket and what needs to be in the ticket has never been more expensive. On April 23, Jim Grundner and I are breaking this down: what’s going wrong upstream, what an automated quality lens for requirements looks like with a live walkthrough of how we’re solving it at Allstacks. Bring your questions. https://hubs.ly/Q049hm-L0 #EngineeringLeadership #SoftwareDevelopment #AIAgents #RequirementsEngineering #DevEx #CodeQuality #TicketQuality[WEBINAR] Your Jira Tickets are Killing Code Quality | Allstacks[WEBINAR] Your Jira Tickets are Killing Code Quality | Allstacks
Experience & Education
-
Visible Flow Consulting
View Steve’s full experience
See their title, tenure and more.
or
Licenses & Certifications
Volunteer Experience
-
Organizer
DevOps Toronto
- Present 13 years 9 months
Education
Bringing together Toronto's technical minds to share, learn and discuss innovation
-
Speaker
Visible Value Stream Consulting
Speaking about:
Flow Engineering
Value Stream Management
Bridging Business and Tech Gaps
Platform as Product - Internal Development Platforms -
Organizer
OpenStackTO
- 4 years
-
Guest lecturer and instructor
York University
- Present 6 years 4 months
Education
Delivering lectures and facilitating interactive workshops to teach lean concepts and practices
-
Guest lecturer and instructor
University of Toronto
Education
Delivering lectures and facilitating interactive workshops to teach lean concepts and practices
Projects
-
Value Stream Management Interoperability Technical Committee
Chair of the VSMI TC working group
-
Value Stream Link
- Present
See projectValue Stream Resources At Your Fingertips
Find everything you need to visualize, measure, and improve
Value Stream Thinking is the key to dramatic and continuous improvement in the post-DevOps era of organizational performance. Software value delivery demands an end-to-end view of value stream operations in order to drive optimal outcomes. This service is dedicated to bringing everything you need to move forward under one page.
One spot to connect to tools, advisors, consultants…Value Stream Resources At Your Fingertips
Find everything you need to visualize, measure, and improve
Value Stream Thinking is the key to dramatic and continuous improvement in the post-DevOps era of organizational performance. Software value delivery demands an end-to-end view of value stream operations in order to drive optimal outcomes. This service is dedicated to bringing everything you need to move forward under one page.
One spot to connect to tools, advisors, consultants, books, guides, presentations. and more. -
Value Stream Management Consortium
-
DevOpsChecklist.com
- Present
See projectCapability assessment within a DevOps framework - highlight your gaps and gain holistic insight on your practices
Honors & Awards
-
Top 10: Great Place to Work 2018
Great Place to Work Canada
https://www.greatplacetowork.ca/en/best-workplaces/best-workplaces-in-canada-less-than-100-employees#2018-best--100-employees
-
C100 Cohort 2015
C100
-
CIX Top 20 2015
Canadian Innovation Exchange (CIX)
Languages
-
English
-
Organizations
-
Toronto Machine Learning Summit
Organizer
- Present -
OpenStack TO
Organizer
- Present -
DevOps Days
Toronto Organizer
- Present -
DevOps TO
Organizer
- Present -
Salesforce Accelerate Program
Member Startup
- -
OneEleven
Member Startup
-
Recommendations received
4 people have recommended Steve
Join now to viewOther similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top contentOthers named Steve Pereira
501 others named Steve Pereira are on LinkedIn
See others named Steve Pereira