What is a fair price for OpenAI custom app dev

Hi all,
I hope it is okay to post it here.

I´m really curious about what a fair price is for the next dev solutions with OpenAI?

  • custom OpenAI(API) app for basic writing stuff
  • custom OpenAI(API) app for analyze and combine data with other for creating own datasets

What do you all think it is a fair price on projectbase or on hour base?

Hopefully someone can inspire me.
Much thanks!


1 Like

This is a difficult question to answer, but my reaction is often neither project or hour based.

I tend to see the world of consulting (AI or otherwise) through the lens of value; a measure of economic benefit the solution provides. Sometimes the economic benefit is time saved. Other times, it could be compressed time-to-market. And yet, another could be the opportunity losses deflected by avoiding distractions.

The hard part in almost every case is determining value. And to be clear, there’s one more layer - you can’t ask for the total value of the solution; your price must be discounted from the total value to compel buyers to “partner” with you to reap the benefits. Indeed, the free market forces suggest you best leave money on the table.

These are really good vectors, but they are each meaningless until you add some value metrics. What will these approaches achieve for the buyer of your services?

Lastly - the cost (not to be confused with your price) is the measure you will compare with your value-based price to determine if it is profitable.

Ironically, let’s use an ice maker as an example. :wink: After all, it was the first robot-like AI tool to enter the home in 1953, the year I was born.

John Gorrie, and the first commercially available ice maker was created by Thaddeus Lowe in 1866. But neither of these machines let owners make ice at home. It wasn’t until 1953 , when Servel began selling refrigerators with built-in ice makers, that ice could be made at home without having to use an ice tray.

  • The ice maker has a clear economic advantage to consumers; it makes, cubes, and dispenses ice automatically. And it repeats this process intelligently.
  • It eliminates the drudgery of filling trays, popping ice cubes out of trays, and generally distracting you from other matters in life.
  • This value is pretty easy to calculate if you know how much each consumer values an endless supply of ice without distractions. Apparently the cost of making ice makers is well below the value they serve since to this day, every refrigerator has one.

Good advice from @bill.french .

I just wanted to second his recommendation for utilizing value-based fees.

While it is harder, and requires a closer relationship with the client, any other type of fees will put you in a race to the bottom.

Alan Weiss is my favorite author on this topic. I highly recommend you give him a listen if you will be providing consulting services in any capacity.

His value-based fees book, and many of his others, are available at my local library. They’re also on audible. And I have hard copies for reference.

Good luck with your ventures into the space!

PS- Since I manage some developer communities, I have been floating the idea of a private group for “AI Practitioners”, specifically for those coming from the developer space. If you’re interested, you can join this wait-list to help steer the possible direction of the new group.

1 Like

Yep - this is the reality of competing based solely on one dimension - price. It telegraphs there are no other differentiators you can offer.

1 Like

Yes. But. Value cannot be measured.

  • Cash flows can be measured.
  • Hours can be measured.

Value is a subjective thing.

Is an automation process that eliminates steps that five people perform in ten hours each per month with a known cost burden, a “subjective” value?

Value is simply an aggregation of the components you listed. Value is certainly difficult to measure and often includes components that are fleeting.

Try putting a price tag on the benefit of five employees eating dinner with their families, or the savings accrued because they aren’t making errors in the automated process.

However, even these curve balls can be factored well enough to provide financial envelopes of probability adequate for decision-making.

If accretive value is unmeasurable, it would be difficult to explain why any automation systems have been commissioned.

Then you’re not working in business. In business, exactly one value matters: Dollars out.
If you can’t show dollars out (or show a path to how your work leads to dollars out) then you’re not in business.

I know this is old, but there is some recent activity so I’ll contribute some of my experience.

I charge 150 an hour to work with the customer. The work isn’t that hard but one of the problems you will get into is the customer wanting all sorts of impossible things. A higher price tag will cutdown on long day dreaming meetings and unreasonable overkill. At first it seems kind of fucked up, but I have saved clients a shit load of time and money by charging this high. Some of these guys want to build LLMs from scratch and all manner of things that are overkill for something like sales bot that maybe talks to 5 website visitors a week. You’re job won’t be just to make the thing for them but also to steer them in the right direction. Their dreams are stronger than your advice, but not their pocketbook.

Another reason I would suggest charging a higher rate is because this trend won’t last but it will make a very small available talent pool, so technically you are worth it. I saw this during the first AI wave, then during the Blockchain wave, and you can see it a little bit in Quantum computing. People were paying million dollar salaries, but unless you get hired, it is sporadic work as a freelancer. Make sure you are charging a larger amount for the gaps in between.