AI for content generation is a long way off

Though AI seems to be a trend that is living up to the hype, content generation is one area where its potential impact is perhaps overblown.

Yes, deep learning is proven to craft more effective email subject lines and other short calls to action (such as in display advertising) but natural language processing is nowhere near good enough to craft long form copy.

Yes, deep learning can be used to manipulate images and even create convincing new ones (see below), as well as create movie trailers it seems, but the training of these networks and the ‘robot hand-holding’ necessary means they are also a long way from proving an autonomous solution.

ai generated images

Images created by generative networks, via paper by Nguyen A, Clune J, Bengio Y, Dosovitskiy A, Yosinski J 

What are content marketplaces?

Content marketplaces are in vogue. They connect businesses with writers, graphic designers, film makers and the like. Though marketplaces are nothing new, the technology they offer is improving and plays a big part in eliminating inefficiencies during large scale content production.

One such marketplace is Quill, which specialises in creating what it calls ‘primary content’, the content that influences consumers at the point of conversion, be it product descriptions or buying guides.

Quill’s cloud platform automates network management, quality control, production and delivery processes. Work can be viewed and edited in the platform, and APIs can deliver the content to a client’s CMS or ecommerce platform. Access to hundreds of freelancers and the automation of bureaucracy such as allocation of tasks and invoicing is what makes this kind of platform a candidate for increasing scale and speed.

For companies with thousands of product SKUs, platforms like Quill are a way to achieve well-crafted content quickly.

Another notable marketplace is Gigster, this time in engineering. Gigster is a software development service which uses more than 700 freelancers to work on projects for corporate giants such as MasterCard and Airbus. One interesting component of the Gigster service is its use of AI to increase the efficiency of its projects.

The company monitors projects to look for patterns that predict bugs or issues with deliverables, assessing activity data across software such as Trello, Slack and GitHub.

Gigster ultimately allows its clients to use a blended workforce of inhouse and freelancers, and to develop projects with much greater speed.

Wider workforce trends

Content marketplaces are part of a wider workforce trend for flexible teams that are able to deal with rapid change, as well as create new digital products and services.

Though much discussion of digital transformation has focused on the need for companies to create cross-functional internal teams, there are also many benefits of maintaining an external network to assist with task-based work. The digital skills gap has been much publicised and marketplaces allow companies to compete in a competitive jobs market.

In Accenture’s recent research into future workforce trends, 73% of survey respondents said that corporate bureaucracies are stifling productivity and innovation. A large majority (85%) indicated they planned to increase their organization’s use of independent freelance workers over the next year (2017).

P&G is one such company that Accenture cites as having recently completed a pilot program using Upwork Enterprise, a freelance management system, with products from the pilot program delivered faster and at lower cost than with conventional methods 60% of the time.

workforce trends

The spectrum of role-based and task-based work, taken from Accenture’s Future Workplace Trends.

The dispersed workforce model – Automattic

There are some companies that have taken the marketplace model to its logical conclusion and whose core team of full time workers is dispersed, too, thereby allowing the company to pursue the best employees wherever they live.

In an interview on Inc.com, Matt Mullenweg, the creator of WordPress and CEO of its parent company Automattic, describes the dispersed nature of its workforce:

“Automattic is a totally distributed company, so everyone works from wherever they are in the world. It could be a coffee shop, it could be their home, it could be a co-working space. We hire people regardless of where they are.

“We now have folks in just over 40 countries. This has been amazing for the company in that we can attract and retain the best talent without them having to be in New York or San Francisco or one of the traditional tech enters.”

Part of making this model work is ensuring effective communication between remote workers. Much like Quill and Gigster using cloud platforms for workflow, Automattic avoids email and uses its own blog theme called P2 for internal comms.

Mullenweg says “I think email is definitely on its way out, between things like P2 and Slack… Email just has so many things wrong with it. I’ve never heard anyone who’ve said they love email, they want more of it–have you?”

He continues, “Imagine if, in your company, instead of email, everyone could post and comment on a blog. Different groups or teams could have their own space on it, but fundamentally everything was tagged and traceable and transparent. That’s kind of what P2 looks like.” 

In summary

There are a number of factors that make freelance and content marketplaces increasingly attractive.

  • A skills shortage means companies cannot always find the right talent to take on full time.
  • Inhouse teams need flexibility, the ability for the team and its skills to wax and wane as projects come and go.
  • Inhouse teams want to avoid bureaucracy wherever possible, using cloud platforms to cut down on admin.
  • Content is so pervasive now and is still a differentiator for businesses both online and off. Compromising on content quality is not an option.

Expect to hear more from marketplaces such as Quill, Gigster, Catalant and Upwork, as the gig economy enables traditional big corporations to innovate in content and beyond.