How to build a business case of artificial intelligence – dan rose he

I have recently surveyed the Danish (main information officials) for their relationship with him and have had some interesting results. One of the results was that one of the biggest obstacles to starting in projects is that building a business issue is difficult. I fully understand the issue and I agree with CIO. Building a business case of it is difficult and if you try to build it as a business tradition, it is impossible.

Building a business issue is about understanding cost and income leaders well enough to work on a model that gives a high security profit within a timely agreed timeframe. When you build solutions or even buy them off the shelf, the whole process turns out to be more challenging than what you will experience with traditional IT projects. In my experience, this is for many lessons learned with difficulties by many in the IT business that of course captures their popular tools and methods but quickly fails. This often turns out that it is ignored as a very immature technology. With the right approach, that I will show you here, you can actually build a business case that makes sense. Technology is ready and at a stage where most businesses can successfully use it. New technology simply requires new access.

Before you move on how you build a business case for him, let’s understand why this is such a difficult task. The reason is simply that everything in it is experimental in its natural form and as a result nothing is predictable. How much data you need, what algorithmic approach will work and how good the result will be is very difficult to know in advance. You can see a similar project, but small changes in the problem, data or the environment will often surprise a big difference. So knowing the exact costs, the results and the road there is not possible.

Cost

Will what will it cost to one? You just can’t know. In the traditional one we try to break down the project into smaller and smaller pieces until each part is such a size that we can easily appreciate the time and costs that enter it. In that process is experimental and we cannot even know the parts in advance.

To combat this problem there is a number of strategies that will make it much easier to control the cost. On purpose I am writing, CHECK and no predictor costs. In the paradigm of it, the cost predicting costs should not be the goal. The goal should be to control it. I’ll be back why this makes sense a little later.

Cost control strategies are:

RECAPITULATION

For years, we have talked about the versatile approaches to it. Some have used it successfully that others have clearly run and stood traditional methods and some unfortunately used it as an excuse not to have a plan at all. Agile access suggests repetitions through projects several times to account for new project lessons and demand changes. Similarly, projects of it should use a repetitive approach to obtain a series of important lessons. Only by making a very fast repetition you just have to get these lessons:

1. You better understand the data. You understand how much effort you are to achieve, how to achieve it and get a meaning how much you will need.

2. You can get a sense of how users react to a certain quality and how difficult it will be

3. You get a good idea of ​​potentially accessible quality.

The last point here should be seen as a stop test. If you do not see a close quality with the acceptable in the first repetition, it is very impossible to see much better results in the near future or with the achievement of much more data or set up a significant investment in the algorithm work. So many projects of it have to be abandoned if the first repetition is not close to a useful solution. In some cases, although this may be just a wrong algorithmic approach. This is where you have to rely on the judgment of technology people.

For the first repetition you can in my opinion start very small using automobile solutions. Automl is the one without coding that can be trained and placed within hours that only need data. There are the pros and cons of being aware. I wrote another blogpost about this here.

FInancIng

I preach a lot about the funds of the moment in him. This is a very effective strategy to control costs. In that point points are natural and project financing must be issued only for each historical moment after a set of agreed criteria is successfully fulfilled. The main point of course would look like this:

Collect data

The first step is to collect a certain amount of data with a certain quality at a certain cost. Collecting, cleaning and preparing data is almost always the most underestimated cost of its projects, so making these criteria for success of very specific first steps is not a bad idea at all.

An important aspect of data collection is the frequency you need to update data. Some projects require only initial or rare data collection and others are required to build a whole data operation that in itself should be a good business case. Many projects die from costly data operations, so consider it early. The trick here is to measure a lot in the process.

Construction model

The next step is to build models. This is in the issue of business not so complicated. This is where technology people should appreciate, but they will do so with great uncertainty and that is as it is. As mentioned before, the first repetition should be as soon as possible and if you do not see the potential for good results after that, you should be willing to stop the project or change technical strategies.

Resort

You should also try to put the patterns of it in a test or environment on stage already in early repetitions. It may seem to overestimate the problem, but the models of it are simply more clumsy to work on my experience than other code bases. Data quantities also make the Dev challenge a little more interesting.

Studies have also shown that up to 99% of the code in projects is all the “glue code” about the current one that makes it work in the given environment. So getting a meaning of this early is also a good idea.

There is also very often a human aspect for setting the model that should be part of the criteria for success. People respond differently to his solutions than other solutions as it is more difficult to understand as secular.

Package your own projects

My latest cost control advice on projects is to merge more projects in a business case. As you can see, the risk of a project he early turns out to be very expensive or not so good, is it there. There is a tendency in it to continue working on projects that have already shown signs of failure as we as people overestimate our ability to improve the situation and we just want to offer something. If we offer nothing, we feel as complete failures.

To avoid this, put more projects in a business case in order to let evil die and the good one time to bloom. You can argue that people just need to be better in his calling when failure is inevitable, but for me changing structures and allowing people to be people is a very high strategy.

The business culture of it

Before crossing the income side, I wanted to add some notes to him and the company’s culture. As I mentioned is that some projects should be closed early. It may seem like a failure but at the right peak this can be seen as a successful NULL risk. Collecting a Ceratin quantity of invalid results is very valuable for a business, especially if done at a low cost. Knowing for sure what a business does not work can sail much easier and plan forward. The only problem is that invalid results are not always acceptable cultural. While it is many of his business strategies to control the cost will not be very easy to implement. So management has a very important responsibility to ensure that the company’s culture supports these approaches.

The same goes for cost control. If there is no culture to control the cost rather than predict that it will hardly be a good experience. He ironacly offers no predictability. So a culture that instead supports the budget or boxing of time is much more effective for it.

Revenue

When someone asks me if a particular problem can be solved with AI, I answer “probably yes” as people are usually on the right track. Question of natural tracking that “How good will he be then?”. The right answer here is “I don’t know”. This is for many people difficult to shallow. People who seek answers here will rarely succeed with Him. Those who may work about that lack of information will be much more likely to be successful, so they must of course go to the business issue.

If the income, value or profit are based on quality, then you cannot calculate the expected profit as you cannot know the results. Even if it is sold at a predetermined price, it is difficult to predict as the approval between users is often based on quality.

Who are you trying to beat?

In addition to using very fast interventions to get an idea about the quality to wait for you also need to be clear what to expect from your own. Don’t try to make a business issue for a perfect one. Make one for a good enough to solve the problem in question. Very often new technology is maintained in gold standards and the expected results will be out of this world. Be very specific here in your communication to avoid this. I also wrote about him here.

Presenting your business issue

Now that you know that you can’t build a business case on the projects as you will have classic IT projects that everyone put on? Not enough. The last challenge is when you have to present the business issue. Your peers who may need to review or accept the business issue usually expect classical cases of paradigm business. So my last advice here is simple – start by presenting the essential principles of him and how this makes the issue of different business. If you get a purchase in your new approach, everything will be much calmer.