How to build and grow AI startups

W H A T I T T A K E S T O B U I L D A N D
G R O W S U C C E S S F U L A I S T A R T U P S
B R O W N E & M O H A N P A G E 1 O F 1 0
Dr TR Madan Mohan and Sharadha V

Introduction
AI technologies are not new. Decades ago, universities and businesses adopted as symbolic
AI techniques. The expert system, rule, or frame-based is the best-known example of
symbolic AI. In recent years, deep learning has emerged as the dominant AI approach.
Tremendous advances in computing power, significant drops in hardware prices, increased
adoption of IT and open source solutions have sped up the image/voice recognition,
prediction, and natural language processing applications. On the business front, we use AI
for three important needs: automation (of repetitive, backend administrative process),
analytics (of what, when, and how), and Customer experience (facilitate conversational
commerce, personalization, DIY, 24/7 presence). AI solutions are the most important
component of the digital transformation of many companies and startups from the USA,
China, UK, Canada, and India are racing ahead to address the needs of industries.
According to Fortune Business Insights, the global AI market was valued at $27.23 billion in
2019 and is expected to reach 266.92 Billion by 2027. CBInsights mentions in 2020 AI
startups raised over $7.4 Billion in funding across 300+ deals from 600+ investors.
In this paper, we analyse AI Startups across verticals to uncover strategic and operational
levers AI Startups use to access complementary assets, sales, and marketing approaches
used to gain market traction, and product development approaches they adopt. We have
accessed AI Startups information from their official records, website, competitive
benchmarking and analyst’s reports, and personal interviews. We have used qualitative
clustering of competitive actions to determine common strategies of AI Startups.
The market for AI Startups is robust, highly
competitive, and offers several advantages for
large incumbent tech firms (Kossuth and
Seamans, 2018). However, AI Startups compete
with the best of solutions with limited data sets
available to them mostly addressing the mid-
size companies across different sectors. In
recent years, Retail, healthcare, and financial
services sectors have witnessed the emergence
of a large pool of Startups, at least in the
logistics and transportation sector. Operational
effectiveness and efficiency, financial and
administrative performance, safety, access, and
customer experience have been the major value
drivers of AI Startups (Agrawal. Et al. 2018,
Cappiello, 2018, Garbuio and Lin 2019).

The common competitive strategies adopted by AI Startups
are:
1) Assisted or autonomous advantage
AI Startups adopt a broad value enhancement strategy,
aligning themselves with incumbent technology solutions.
Assisted or Autonomous are two broad approaches used
extensively. Assisted AI solutions wrap around existing
processes/solutions (sometimes without replacing the
human element) and improve performance and
productivity. The focus here is to help client organizations
accomplish their tasks better. Assisted AI startups are
careful not to disrupt dominant incumbents and align to
work with all the majors. Hence, compatibility and
integration with existing platforms become a necessity.
The key here is also to preserve or enhance the dominant
platform’s revenues these could be an agency management
system or a healthcare medical records application
wherein the existing applications may suffer from limited
access, operational performance, and service experience.
Assisted AI startups can successfully pursue a “built to sell”
strategy by bringing out valuable solutions to elicit interest
from the market and enjoy a profitable exit. They endow
these AI startups with deep domain knowledge, not just
technical skills to identify white spaces, incubate solutions
in early client environments, and scale up the solution.
Assisted AI startups can also choose to be biased with a
larger incumbent to ride on its market share. On the front
of the resource, Assisted AI startups can pursue
incremental capabilities in people and systems and de-risk
themselves by working with large established players using
open innovation contracts.
How AI Startups evolve and grow
AI Startups despite the difference in sizes, investment or industry focus have common
approaches to the way they have evolved from ideation, incubation and growth stage. We
have deliberately avoided investment and IP related strategies, but rather focused on
product, market, partner and innovation areas.

Exdion Solutions has developed AI solution that completely automates the entire policy
checking process, taking only a few seconds to run through a policy document and
checking for E&O errors on policy documents. Evaluator offered by Streamline Health is
another solution that checks the accuracy of data entered in an existing billing system in
case of any errors and alterations created by copying and pasting between systems. X. ai
offers an AI intelligent personal assistant that updates or schedules meetings and more for
early-stage startups to enterprises. It integrates easily with programs like Outlook, Google,
Zoom, Office 365, and Slack, along with continually learning from every data scenario
provided.
Autonomous AI solutions work
independently of human intervention.
Autonomous AI startups need
“resource munificence” both in terms
of skills, model building, testing, and
implementation. Prudent management
of raising funds and product
development is essential. Market
signaling on innovations is key to
attract market interest. Autonomous
AI startups may need to play a co-
opetition strategy by working with
competitors, to meet the need of the
sponsoring platform or the OEM. They
might also need to invest in resources
upfront to take part and manage
standard-setting and engage with
certification and regulatory
authorities. Autonomous AI startups
prepare themselves for a “build to
scale” model, open to investment and
partnerships from large incumbents.
Pony.ai and Aeye companies are
autonomous AI startups that are
building vision algorithms to guide
autonomous vehicles.
2) Co-create to access data and sharpen offering:
AI startups work closely with early adopters to get insights about the product and market
fit. For large incumbent players, AI skills short-supply, limited experience of building
scalable solutions, and with “small bet, no regret” innovation outsourcing approach is a
viable approach to grow revenues and drive efficiencies. Sourcing and labeling data sets are

expensive and hence AI companies use open innovation contract to work with large
established players to show the technology and algorithmic efficacy, explain how the
solution will work with the company’s data and how long it will take to train the AI solution
on the company’s data. AI startups are also using the open innovation contract to assess the
integration of AI solutions with existing systems, compatibility with the technology stack,
and product roadmap alignment with future needs of the industry. AI startups are using
open innovation to assess legal and reputation risks of their solution, the transparency and
applicability of their algorithms, the provenance of data used, and any data sample bias.
3) Unbundle solutions
Successful AI Startups initially conceive
and attempt to develop a robust and all-
inclusive solution. However, proof of
concept (POC)’s and early customer
adoption insights make them realize the
folly of pursuing a full-blown solution
strategy. Industry maturity, investment
risks, and change management challenges
tip the AI experimentation to smaller
chunks of the solution. Unbundling offers
two advantages to AI startups. Customers
can choose to adopt AI solutions in
chunks so that change management risks
can be mitigated and adoption efficiently
managed. Unbundling also helps
customers pay for modules they will
adopt, resulting in lower budgeting
complexities.
Antworks, the primary product is
ANTstein, an AI-based RPA software
product designed to understand
structured and semi-unstructured data
types such as handwriting and forms.
Antworks has neatly unbundled its
solution to cover ANTstein Square,
Process Discovery, Image Enhancer, Auto
Indexer, CMR, QueenBot (The QueenBot
incorporates: Cognitive responsiveness,
Intelligent digital automation, and Multi-
tenancy, Email Agent, Accounts Payable
as a Service and Optimiser).

4) Move to Value-based pricing
Pricing is a thorny issue for AI companies. Pricing
has traditionally revolved around cost and
competition. AI startups realize pricing is cross-
functional (from finance, operations, and others)
and value-based. AI startups offer proof of
concept as a fixed fee or free, but for the license
to deploy and use has resorted to value-based
pricing. Moving away from pricing based on deal
size, AI start-ups are increasingly resorting to
assess the perceived value the customer is likely
to derive from the adoption of a solution.
As they scale up, AI startups realize smarter value-
based pricing works best, especially since their
solution is modularised. AI startups also realize
the fallacies of tying up pricing models on a per
person basis and move to a more hybrid value
model involving several documents and accounts
searched.
OliveAI is yet another good example of how AI startups benefit from unbundling. The
company efficiently has un-bundled each of its solutions ranging from audit, industry
benchmarking, and knowledge data modules that consumers can choose from based on
their need. Unbundling also helps with pricing, as a customer can choose to pay for only
the module they wish to use.
5) Horizontal and Vertical expansion
AI solutions can be vertical or horizontal. Vertical AI startups solve problems by
expanding their extraction, generation, and recommendation capabilities in a specific
industry. These Startups expand either by deepening their offerings within the same
domain or move to apply the same solution to other industry problems. Vertical AI
implements more sophisticated noise reduction techniques to improve extraction or
improve the algorithms that are to apply to similar problems within the same domain.
Often, their expansion is also on the value of the data the solution is churning for
predictive intelligence. Viz.ai AI-powered app helps identify suspected large vessel
occlusion (LVO) stokes. While its Viz ICH product is a related market technology
expansion, Viz RECRUIT is an alert system for the research team about potentially eligible
trial patients and facilitates enrolment. Horizontal AI Startups fit a lot of use cases and
their expansion into other industrial verticals happens becausethe underlying business
processes are most common or the functional knowledge is portable to another function
for quicker decision-making.

http://www.browneandmohan.com
6) Data-centric to templates, to fractals
Majority of the AI Startups start building
models based on data and realize the benefits
of managing information at business template
level as variety could be better controlled.
Initially managing the right balance between
noise and extraction would be a challenge. As
data sets grow larger and ask of business
insights increases, performance of the
applications based on AI neural networks
suffer. AI Startups recognize multiple options
to address this issue. AI Startups realize
while businesses use a different application
to access and use data, the business
templates used across different applications
is limited and standardized. Moving from data
to business template based extraction and
matching is a possibility. Using fractal data
for the existing ML algorithm or use of fractal
models to design new machine learning
algorithms are other available options. AI
Startups attempt fractal analysis wherever
characterization of noise is good. AI Startups
realize the challenges of working with fractal
analysis using attractors come with
limitations on automatization front. Limited
resources with a good grasp of chaos theory
and the need for an extensive amount of time
to curate applications pose challenges.
Eugenie.ai and Cuddle.ai started with data
sets and realised the benefits of using
patterns.
Uptake is an Industrial AI and IoT platform that helps companies manages their hardware
assets. Uptake realized many of its customers have dealer partners who sell and distribute
products. It has now expanded to support sales and customer service by building the right
use cases. Affectiva is an AI start up offering Emotion as a service solution that developed a
face/emotion recognition feature for Media Analytics to understand their consumer’s
reactions towards certain products. The solution was deployed across the same vertical
with minor changes targeting media from advertisements to movies. With minor changes in
its technology, they deployed the solution use of across industries from biometric usage to
transportation to analyse in-cabin emotions.

7) Limited data to own Data Lake
Limited data is the “Achilles heel” of most AI startups. They may start with limited client data
to establish a baseline, reach the required accuracy and precision, and limit loss. Smart AI
startups gather data across implementations and store it at the leaf level in an untransformed
or nearly untransformed way. Some AI Startups use the industry data to offer “audit and
benchmarking” services so that client organizations can assess their current situations and
scope of improvement. The additional services open up an additional revenue stream for AI
Startups. Olive AI’s, Deep Purple gathers this contextual information, allowing Olive to find
new connections and opportunities to improve her work. By applying far-ranging insights that
Deep Purple amasses, it allows providers to discern patterns and relationships among diverse
data types and sources.
8) Hone partner play
AI startups rely on partner ecosystem to reach new
customers and mine existing customers. Partner
management is a key component of their go-to-market
strategy. From tiered value-added resellers to white
labelling, AI Startups exploit various forms to engage the
market.
Zesty.ai, an insurance tech startup uses artificial
intelligence to understand the impact of climate risk. The
company partnered with IBHS or the Insurance Institute
for Business & Home Safety; the research provided by
IBHS along with Zesty.ai’s extensive historical loss data,
will enable the development and growth of AI-powered
risk models for the insurance industry. ClimaCell is an
all-in-one weather intelligence platform that predicts
and automates weather challenges. It has partnerships
with Google Cloud to provide access to weather
forecasting models in geographies that currently lack
such services.
Patient pop, a startup that uses Ai for handling marketing
is currently on the marketplace platform of Athena, a
leading healthcare records provider. With this, Patient
Pop now has access to a wide range of resources offered
in the platform and the wide range of access to other
clinics and hospitals that use Athena’s health record
system.

9) Change management and adoption
AI startups, both assisted and autonomous, need to invest in strong organizational change
management and communication initiatives before, during, and after AI implementation. AI
startups need to educate client organizations early for a smooth transition and tailor-make
adoption to multiple stakeholders in each client organization. AI startups provide
expectation workshops where stakeholders are equipped with baseline knowledge of the AI
solution’s workings, changes it might bring to the workflow and benefits they can expect
from these changes. This may include everyone from administrative resources to key
solution users. AI startups assess adoption progress, track against goals, and quickly
address challenges to stay on course to meet the goals.
10) Managed services
Most AI startups realize many of them may
not be mature enough to manage AI
interventions on their own and continue to
prefer engaging the AI solution provider.
While this may pose scaling up challenges,
especially for a pure play product business
plan, the option provides additional revenues
and access to change/improve the solution.
An example of this would be a healthcare Ai
streamline health offering audit and
compliance solutions, once the solution
identifies where the client’s compliance
stands, the services offer where
improvements are needed and further
advances it with what they could do and
correct it.
Conclusion
AI startups, whether assisted or autonomous, will continue to grow in numbers shaping many
industrial and service sectors. Skills will continue to be a major challenge, and AI startups
must explore hybrid models to explore and exploit skills and capabilities of their ecosystem,
including institutes of higher learning. Advances in computing and AI techniques will continue
to create opportunities and challenges to the growth of AI startups. Focus on building
relationships, agility to reconfigure their offerings and pricing, and remaining open to change
their business strategy is imperative for all AI startups to survive and grow. How quickly AI
startups can instill sense and respond capabilities will determine their relevance and business
performance.

Selected Bibliography
Agrawal. A, Gans, J and Goldfarb, A. (2018) Prediction Machines: The Simple Economics of
Artificial Intelligence, Harvard Business Review Press, Boston.
Cappiello, A. (2018) Technology and the Insurance Industry: Re-configuring the Competitive
Landscape, Palgrave Pivot.
Garbuio, M. and Lin, N. (2019) Artificial intelligence as a growth engine for health care Startups:
Emerging business models. California Management Review, 2019, 61, 59–83
Kiron, D and Schrage,M. (2019) Strategy For and With AI, MIT Sloan Management Review, Vol.
60, No. 4, 30-35
Kossuth, J and Seamans, R. (2018) The Competitive Landscape of AI Startups, HBR Online Print,
December 21, 2018.
B R O W N E & M O H A N P A G E 1 0 O F 1 0
Browne & Mohan white papers are for information and knowledge update purpose only. Neither
Browne & Mohan nor its affiliates, officers, directors, employees, owners, representatives nor any of
its data or content providers shall be liable for any errors or for any actions taken in reliance thereon.
© Browne & Mohan, 2021. All rights reserved Printed in India

How to build and grow AI startups

More Related Content

What's hot (20)

Similar to How to build and grow AI startups (20)

More from Browne & Mohan (20)

Recently uploaded (20)

How to build and grow AI startups