Quais são as melhores maneiras de persuadir um colega de trabalho a usar uma solução algorítmica?

I will do my analysis to be sure that algorithmic approach will add significant incremental value & resources are available to build the algorithmic solution. Then coworker and myself will have a candid discussion about current approach's performance & it's pain points. For example, suppose revenue forecast & it's scenarios are created manually & I believe we should switch to AI algorithm. Coworker & myself with discuss manhours spent currently, current performance, errors observed, delay in meeting timelines, criticality etc. If current approach is falling short to meet it's objectives & there are significant challenges with current approach, then we will align to explore the algorithmic approach.

The basis of explaining and solving a problem is the definition of objective metrics for measuring the performance of the solution. This sets the stage for a productive-objective discussion regarding different proposed solutions. The objective metrics enable the comparison between baseline or existing solutions with the new improved ones.

Trying to bring a solution for something people don’t believe it’s a problem is a lost cause. I like to think of algorithmic solutions as a enabler that removes pain points: - this process is tedious; - this process is inaccurate; - this process is time consuming; - etc. Let’s take the example of formatting code. Problem: people format code differently. It cause issues when merging the code in a large team. The team is frustrated with the process of comparing code that appears to have changed but it’s just formatting. Key points: you have a large team struggling with merging code. Algorithmic solution: apply a code formatter to all code before pushing the code to the repository. Outcome: the team is less frustrated.

One thing I've found helpful is to explain how the algorithm that I've taken is more appropriate and also giving them a chance to prove theirs if more efficient

Equally important is actively listening to your coworker's perspective and feedback. Engage in a collaborative dialogue to understand their insights and experiences. This not only enriches your analysis but also fosters a sense of shared understanding. By establishing common ground through thorough communication, you not only build rapport but also lay the groundwork for trust. When your coworker feels heard and understands the depth of the problem, they become more receptive to your proposed solution. This collaborative approach ensures that your solution is not only effective but also well-received within the team, fostering a positive and constructive working environment.

Once we have agreed that the problem exists and we should look for alternate solutions, then I highlight the benefits of algorithmic approach. Efficiency gains through manhours saved, increase in performance, ability to scale, reduction in errors due to automation, ability to automatically adjust with changes in business environment etc. Then it's important to do a POC to quantify the real improvements in accuracy, time saved, speed etc. This will convince myself and coworker that it will really work, before we take it to higher management for funding or approvals.

Certainly! After clearly defining the problem, it's essential to showcase the superiority of your algorithmic solution over the existing method. Emphasize the advantages and positive outcomes your solution brings, whether it's enhanced accuracy, speed, scalability, or simplicity. Highlight the value and impact of your solution, emphasizing how it addresses the identified problem effectively. The goal is to demonstrate not only the efficiency of your solution but also its practicality and real-world impact, ensuring that your coworker can appreciate the value it adds without getting bogged down by unnecessary complexities.

The benefits of a new solutions should be compared in objective manner with already defined metrics which enable innovation and avoid overengineering. The rule of thumb is that a complex solution is not a solution.

Discounting other's ideas is easy. I believe this is because it takes effort to completely understand the solution and its benefits. Therefore unless you explain the benefits, it's difficult to get a complete buy-in on the plan. So once it's done, things start moving quickly.

It's always important to look at the problem from all the views including high level and deep dive approaches. Be ready for constructive feedback and objections. Once the entire picture is clear, emphasize on a healthy discussion with the coworker, discuss on the values such as how your solution solved the problem in less time and space. Discuss how the solution will benefit in less resources and most importantly try to maintain the decorum of the community.

I listen to my co-workers concerns and respond to them honestly. My co-worker can have questions about time needed to build the algorithmic solution, cost, resources needed, resource availability, expected performance improvement with algorithmic approach etc. If I feel that these concerns will be concerns of other people also, then I will try to address them when my team is building the algorithmic solution. Also the points raised by the coworker can be raised when we go to senior leadership for their approval and or funding. So I will prepare well to answer these concerns before our meeting with senior leadership.

Absolutely, when addressing your coworker's concerns or objections to your algorithmic solution, approach it with understanding and transparency. Crucially, don't dismiss or ignore their feedback; instead, view it as an opportunity to refine and improve your solution. Use their insights to bridge any gaps or address potential shortcomings, fostering a collaborative environment where concerns are genuinely considered and integrated into the solution's development. This approach not only strengthens the credibility of your proposal but also demonstrates a commitment to continuous improvement and a shared goal of addressing challenges effectively.

It is always a good idea to create a working doc of your change and share it with your coworkers rather than handling each query independently as the queries might get quite repetitive. At the same time, the concerns/questions of the coworkers should be answered with respect irrespective of how basic you might think they are. The goal should be to make the transition as collaborative as possible

Recognise and accept your coworker's reasonable worries about the algorithmic approach. Give truthful responses about cost, time, and hazards. Show how the solution solves ethical and societal issues. Use their comments to improve the solution collectively, ensuring it meets their needs and preferences.

It’s safe to assume one person’s concern is the same concern for many others. So addressing one concern is addressing multiple folks and this will allow further follow-up and prompt feedback. This will also help others to understand and get on board. Plus you get to know more folks which is a great thing to do.

Murder those who refuse to do as you say. This has been an effective technique used throughout history. Beware, this behaviour is illegal in many jurisdictions.

Let me take the case of my coworker being from the stakeholder team. In this situation, I feel it's mandatory to have my coworker be involved in building the algorithmic solution. If my coworker is is busy, I request my co-worker to ask one of his/her experienced team member to be involved. Stakeholders understand their business best, they know the data and understand what their clients need. Hence we involve them in designing phase, and review the solution with them, to make the solution more relevant. It increases the confidence and ownership of the co-worker in the solution. It also gives the co-worker due respect and make them feel empowered.

Involving anyone in process works better because person feels part of the project and also more confident to provide counter arguments as well as understanding why a particular algo or approach was used. Involving them in process also gives them chance to know what did not work, why and what was the better solution. Once coworker is comfortable in above, it is higher chance he/she can grasp and think about expected algorithmic solution

While addressing concerns it is a good idea to identify component owners. The work required may span multiple teams and this would be a good idea to have owners address specific issues. Encourages team effort resulting in smoother project execution and learning from others.

I think this can be a best way out to get the stuff done. It may happen that if the coworker has a criticising nature, he/she may not agree on the approach. They may underestimate your efforts. Adding them to your squad, gives them a feeling of involvement, and there are more chances of them being on the same page. They would acknowledge the hurdles and can give you feedback and suggestions which may differ from your perspective.

It’s often that my coworker is also potential stakeholder who decides whether to move from a manual to algorithmic system. I respect the final decision of my coworker/stakeholder. It’s the stakeholder who takes ownership of the final outcome and hence is entrusted with the final decision. I don’t think that this is where my professional relationship closes with this coworker. As time progresses the coworker probably will have more confidence in algorithmic solution and build trust upon me if I am able to show him/her successful user stories. It has resulted in him/her asking me to help to move to algorithmic solution even after couple of years. Also there have been 2-3 times when they recommended me to other potential stakeholders.

If a solution meets all objective metrics set in the beginning, even if it's not an algorithmic solution, it deserves respect and acceptances. Simple solutions are always better.

It is crucial to respect coworker's decision, even if they choose not to proceed with the algorithmic solution. Acknowledging their perspective and maintaining open communication to understand their concerns or considerations. This will fosters a positive working relationship and allow for future collaboration on finding the best solutions for the team.

I would add that sometimes we have a misconception about the problem/solution. When getting into this part "respect their decision", it is important to think if we are missing something on the issue.

Don't let others dictate you how to do your job and vice versa. It's good to suggest and drop hints, but being dismissive or noncooperative is unethical.

One thing to ask yourself is if you really need to use an algorithmic solution? Are the pros really that big? Engineer who owned that algorithm might leave but the code is what lives. Simplicity and readability should be the utmost important metrics that should be taken into account after performance. For e.g. Should we add a hard to read and maintain distributed algorithm for a 3-4% performance gain? NO!

One thing to answer yourself is: why persuade the coworker to a specific algorithm? Does it have a better space/time complexity? Is it an outcome of the latest research? Is it just because you "like it". Choosing a solution has pros and cons, and unless there is no explicit benefit of using a different algorithm the coworker should be free to use whatever they like. Why? It builds their confidence, experience, and a sense of fulfillment. An example might be: Coworker wants to use a simple O(n^3) algorithm, but a more complex O(n^2) algorithm exists. Should you use a faster, more complex solution? Depends. For small data, the answer is usually "no", simplicity beats speed here, especially if this is not the system bottleneck.

Showing by doing. I suppose that the most impactful persuasion is to demonstrate them. Say your colleagues spend 30 mins to perform the same task that you spend 3 mins and deliver similar results.

We often forget to question ourselves under a real time-resource constraint. If something doesn't go our way, it necessarily doesn't have to. We should be willing to drop an algorithmic solution, even if it is the best fit and only one criteria (can be acceptance from co-worker, can be anything else) is causing a shutdown, its better to wait for iron to get hot so we can hit the hammer at the right time.