Join the Community

22,037

Expert opinions

43,964

Total members

420

New members (last 30 days)

182

New opinions (last 30 days)

28,682

Total comments

Join Sign in

The software ownership dilemma

12 August 2024 2 comments

Erica Andersen

Marketing

smartR AI

Software ownership has undergone an interesting evolution. Traditionally, software producers retained ownership, through licensing or the use of Software as a Service (SaaS). However, the AI era provides an opportunity to bring about a significant shift.

A key difference in the AI era is the increasing customization of software for individual users. The ability to retrain models with client data allows organizations to capture their unique tribal knowledge. This customization offers significant advantages over "off-the-shelf" models like ChatGPT.

Organizations typically prefer to keep their data and information secure, away from potential competitors or third parties. Ensuring the safety of data already is a crucial concern for companies in this AI era, but will be even more so with this customized information and data When models are highly customized for each enterprise, who will ultimately own these models? The market is still undecided, but innovative startups like smartR AI and SeerBI are experimenting with transfering model ownership to end-users.

Transfering software ownership to consumers offers several advantages. First, this approach provides end-users with a higher level of security and control that has been lacking in traditional software ownership models. Second, they can choose where to deploy the model. Third, they have assured access to the model, regardless of the vendor's circumstances. Fourth, as the software becomes an enterprise asset, the model can be sold or licensed to third parties, representing intellectual property for the organization.However, one drawback of owning the software is the question of how the enterprise will support and upgrade it. To address this, companies can enter into multi-year support contracts with vendors to handle software maintenance and upgrades.

This new approach to selling software provides enterprises with a fresh perspective on their intellectual property, and the role of software in their core value proposition. Bespoke AI captures significant enterprise value and changes the way software is owned, offering organizations greater control over their digital assets.

Written by: Dr Oliver King-Smith is CEO of smartR AI, a company which develops applications based on their SCOTi® AI and alertR frameworks.

External

This content is provided by an external author without editing by Finextra. It expresses the views and opinions of the author.

1767

Report

Channels

/artificial intelligence /security

Artificial Intelligence and Financial Services

Join group

366 opinions 89 members 17 November 2024

Comments: (4)

Ketharaman Swaminathan Founder and CEO at GTM360 Marketing Solutions

13 August 2024

COTS and SAAS come with very little data out-of-the-box. Almost all the data that's inside them in a live installation is entered by customer (or their implementation partner) during the implementation. While they come with a lot of features out-of-the-box, most midsized to large enterprise implementations involve extensions / customizations specific to the customer. STILL, all large COTS and SAAS vendors have somehow managed to retain ownership of the entire software, including customer-specific data and extension / customizations. While the vendor cannot make one customer's data available to another customer, it can - and often does - merge most extensions / custoizations into the future version of its product and thereby makes them available to other customers. This is despite almost all customers asking for joint ownership. Since I've been in enterprise software sales for decades, I know exactly how vendors manage to wrangle unilateral ownership but that's not the key point in the current context.

In the case of GenAI, the out-of-the-box software contains a lot of data on which the models are trained by the vendor - data that's not supplied by the customer. IMO it'll be way easier for GenAI vendors to wrangle total ownership of data.

In other words, I'm predicting the exact opposite of your prediction. Once again, vendors won't be able to make one customer's data available to other customers.

1 Like

Report

Erica Andersen Marketing at smartR AI

Author 15 August 2024

Thank you, Ketharaman, for your insightful comment. Your perspective, rooted in decades of enterprise software sales experience, is invaluable.

I agree with your assessment of how COTS and SaaS vendors have historically maintained ownership of their software, including customer-specific data and customizations. Your point about vendors merging these customizations into future versions is astute and often overlooked.

Regarding GenAI, I concur that the initial models contain vast amounts of vendor-supplied training data. However, I believe the situation becomes more complex when we consider model specialization using proprietary company data.

Unlike traditional SaaS products, fine-tuning an AI model with company-specific data fundamentally alters the model itself. The weights of the neural network are adjusted, effectively embedding aspects of the proprietary data into the model's structure. This creates a unique challenge: the data becomes inseparable from the "software" (i.e., the model).

This inseparability is where I foresee potential issues. If a vendor were to use this specialized model for another client, they would inadvertently be transferring knowledge derived from one customer's proprietary data to another. This scenario is fundamentally different from traditional software customizations, where features can be more easily isolated and repurposed.

That said, I appreciate your prediction that GenAI vendors will find it easier to maintain total ownership. It's certainly possible that they will develop novel strategies to navigate these complexities.

Perhaps the solution lies in developing more sophisticated methods of model specialization that can clearly delineate between base knowledge and customer-specific insights. This could allow vendors to maintain control while ensuring customer data privacy.

I'm curious about your thoughts on potential solutions to this dilemma. How do you envision GenAI vendors balancing ownership rights with the need to protect customer-specific knowledge?

Report

Ketharaman Swaminathan Founder and CEO at GTM360 Marketing Solutions

20 August 2024

Here's my pointwise reply:

"This creates a unique challenge: the data becomes inseparable from the "software" (i.e., the model).". Bingo! "Inseparability" is EXACTLY how vendors will pitch the need to own all data (and features)!

"If a vendor were to use this specialized model for another client, they would inadvertently be transferring knowledge derived from one customer's proprietary data to another. This scenario is fundamentally different from traditional software...". I'm NOT so sure of that. From time to time, SAAS vendors have been nailed for transferring knowledge derived from one customer's proprietary data to another e.g. Carta, Oracle, Salesforce.

As in the past, vendors will make a lot of song and dance during presales about "balancing ownership rights with the need to protect customer-specific knowledge" but the proverbial footnote #3 in page #42 of their contract will likely vitiate that when the rubber hits the road.

Report

Erica Andersen Marketing at smartR AI

Author 22 August 2024

Thank you for your informed opinon. It's always good to hear back from others in the community with AI insights.

1 Like

Report