Where is the line? Isn’t the goal to get paid?

Having spent the better part of 2017 shifting the company focus to Deep Learning consulting services, we experimented with various approaches to explain and show to potential clients what it is we do and how we do it. As an AI consultancy, we have the same growing pains as any other shop, plus the volatility of technology that changes daily.
The motivation
Many of our potential clients came to us with one thing in common: they had all tried another AI shop before us. "We spent $250k for a report", lamented one CEO. "They disappeared for nine months and we got lukewarm results at best", said another.
This is when we realized that there are a lot of charlatans and would-be’s out there, and it’s up to us to show that we can walk the talk. The promise of money attracts all kinds of people; the word-slinging suit and the smoke-and-mirrors magician are definitely some of them.
That isn’t to say that there are incredible teams out there; rather, demand has far exceeded supply, which is skewing the power in the relationship to whoever puts a .AI on their domain name.
A line in the sand
So now that we’ve established that there might be a need for a demonstration of skill, how much work can a consultant do without a signed contract? Below are our in-house guidelines that we’ve developed to keep the lights on and for everyone to be able to buy groceries.
Approach 1: The 24-hour prototype
The simplest approach is this: show the client that you can do something related to their field. There are a lot of public datasets available, so there’s no need for the client to provide that for you, depending on the scope of work.

Approach 2: The sample data
Alternatively to the prototype build based on open data, you find the most direct insights from within the client data itself.
"Show me what you can do," said a client after downloading a bunch of PDFs to our servers. They selected a handful of reports of interest, and we showed them how entities grouped together in terms of cooccurrence. Done and done, signed contract.
Approach 3: Solution architecture
If you were to build the ideal solution, what would it look like? After a few dozen projects under the belt, we can now go to a client and discuss what a deployment would look like: technologies, stage-gated phases, and lessons learned. After all, aren’t your clients hiring you for this exact reason?
Since our secret sauce lies not in the Lego Blocks that we choose for the project, but rather how we put them together, this solution is our most successful approach to date (total effort vs. closing rate).
Lessons from the past
Although we honed these lessons over time, they were also forged in the fire that is applying knowledge for money. Where there is money, there is greed.
IP theft
"Aren’t they just gonna steal your idea?" A very good question. Let me answer that with another question: If they were gonna screw you over, isn’t it better to find that out at the beginning of the would-be relationship?
Let them take a snippet of what would be an otherwise disastrous situation. Feed their inner kleptomaniac, and then block their phone number.
Pick the right client
Difficult clients are difficult from the beginning to the end of the engagement. If they start nitpicking your prototype, then they either 1) do not know what a prototype is, as opposed to a complete product, or 2) they are trying to get more for less.
An easy client will be clear:
- "I want that. How much will it cost?" My favorite. Make sure that you understand your own pricing logic through and through. Here’s my CTO’s opinion on the matter: https://towardsdatascience.com/how-to-price-an-ai-project-f7270cb630a4
- "Can we install it on X, or connect to Y?" Clear requirements. A good indicator of very low scope creep.
- "We can start with X and grow it from there." Limiting the scope. Love it.
Walking away from problems, not money
Alternatively, here are some of the red flags:
- "That expensive, huh?" Yes. You planned your time according a particular hourly rate. They may actually not have the money.
- "Can I take a peek at the source code?" Nope! Your secret sauce is your lifeblood. You may share it with your client once they paid, but never, ever before.
- "Just make it pretty and we can sell it as is." Without functionality? And what will the client tell you, now that your name is to blamed?
- "If you build it out just a little more, then I can sell it." This is a tough one, because it could be true. In this case, ask for timelines, budgets, and champions on the client side. If none of these exist, bow out.
As with anything, your mileage may vary. Trust, but verify.

In short…
Remember that the goal of these prototypes is to showcase skills so that you may get hired, and not to do all of the work upfront.
Happy Coding!